economy-wide modeling: benefits of air quality ... · quality into economy-wide modeling, this...
TRANSCRIPT
1
Economy-Wide Modeling: Benefits of Air Quality Improvements White
Paper
September 22, 2015
Prepared for the U.S. EPA Science Advisory Board Panel on Economy-Wide Modeling of
the Benefits and Costs of Environmental Regulation
This paper has been developed to inform the deliberations of the SAB Panel on the technical merits and
challenges of economy-wide modeling for an air regulation. It is not an official EPA report nor does it
necessarily represent the official policies or views of the U.S. Environmental Protection Agency.
2
[This page intentionally left blank.]
3
Table of Contents
1 Introduction ............................................................................................................................. 4
2 Air Quality Benefits .................................................................................................................. 5
2.1 Air Quality Benefits Traditionally Included in Benefit-Cost Analyses ........................................... 5
2.2 General Framework for Benefits Valuation .................................................................................. 6
2.2.1 Temporal Dynamics of Health Impacts ............................................................................... 12
2.2.2 Income adjustments ........................................................................................................... 12
2.3 Conducting and Presenting the Analysis ..................................................................................... 12
3 The Benefits of Air Quality Improvements in CGE Modeling ................................................ 14
3.1 EPA CGE Benefits Modeling – Benefits and Costs of the Clean Air Act from 1990 to 2020 ....... 14
3.1.1 Changes in Medical Expenditures ....................................................................................... 15
3.1.2 Effects on the Labor Force .................................................................................................. 15
3.1.3 Potential Analytical Limitations .......................................................................................... 17
3.2 Outside Studies Modeling Direct Health Impacts of Air Pollution in CGE Models ..................... 19
3.2.1 Potential Analytical Limitations .......................................................................................... 24
3.3 Non-Health Environmental Feedbacks in CGE Models ............................................................... 25
3.4 Benefit-Side Tax Interaction Effects ............................................................................................ 29
4 Additional Factors Affecting the Incorporation of Benefits into CGE Models ...................... 31
4.1 Valuing Mortality Risk Reductions .............................................................................................. 31
4.2 Health Risks to Children .............................................................................................................. 34
4.3 State Dependent Utility Functions .............................................................................................. 37
4.4 Indirect Impacts on Human Health ............................................................................................. 41
4.5 Air Quality and Labor Productivity .............................................................................................. 42
4.6 Material Damages from Atmospheric and Deposition Effects ................................................... 43
4.7 Air Quality and Productivity in the Agricultural and Forestry Sectors ........................................ 43
4.8 Spatial Resolution of Benefits ..................................................................................................... 44
4.9 Spatial Sorting and Welfare Effects ............................................................................................ 46
5 Concluding Remarks .............................................................................................................. 47
References .................................................................................................................................... 49
4
1 Introduction
Economy-wide analyses of tax and trade policy typically consider the benefits and costs of those
interventions simultaneously. This is, in part, because of the important interactions between the two that
help define the overall net impact on the economy and society. In these cases, it may be more natural or
straightforward to consider the economy-wide benefits of policy interventions due to their market nature.
However, the typical non-market nature of benefits associated with improved air quality does not imply
that they have no impact on the economy or do not interact with the costs of realizing those
improvements. While it has become commonplace in applied benefit cost analysis to assume that benefits
are additively separable in households’ utility functions and therefore do not affect other household
choices, this assumption is at odds with much of the applied environmental economics research on the
benefits of improved environmental quality. Many estimates of the benefits associated with
environmental policies leverage changes in market activity that complements household consumption of
non-market goods and services as environmental quality varies. For example, households’ willingness to
pay for environmental improvements are commonly estimated based on the way spatial and/or temporal
variation in environmental quality affects housing prices due to its influence on housing demand
(Palmquist, 2005). Similarly, studies leveraging observed housing market behavior in response to variation
in environmental amenities, have estimated households’ willingness to pay for improvements in health
(Gayer et al., 2002). Benefits from improvements in environmental quality are also routinely estimated
from consumers’ willingness to trade-off other consumption for improved environmental quality in their
recreation decisions (Phaneuf and Smith, 2005). Furthermore, improvements in environmental quality
have direct effects on economic markets through improved labor and agricultural productivity (Zivin and
Neidell, 2012; US EPA, 2011), amongst other direct impacts. Therefore, the benefits of improved air quality
may have important economy-wide impacts and may interact with the cost of achieving those
improvements, such that including both benefits and costs in an economy-wide analysis of some air
regulations may be important for assessing the net welfare effects. To this end, the chapter on economy-
wide computable general equilibrium (CGE) models in the Handbook of Environmental Economics states
“if the [CGE] model is to be used for evaluation of policies, it should be capable of quantifying both the
costs and the benefits of the policies in question” (Bergman, 2005).
Introducing the role of the environment in production and household behavior at an economy-wide scale
is an area of applied analysis that is relatively new. However, the suite of studies that have partially
incorporated the impacts of environmental quality in economy-wide CGE have found substantial
economy-wide impacts of improvements in environmental quality (Matus et al., 2008; Carbone and Smith,
2013). In its analysis of the benefits and costs of the Clean Air Act Amendments (CAAA) from 1990 to 2020,
EPA found including only part of the human health benefits associated with improved air quality, along
with the compliance costs, in a CGE model dramatically changed the estimated economic and welfare
impact of the regulations (US EPA, 2011). In its review of that analysis, the Advisory Council on Clean Air
Compliance Analysis stated that inclusion of benefits in the economy-wide model specifically adapted for
use in that study “represent[ed] a significant step forward in benefit-cost analysis.” However, serious
technical challenges remain when attempting to evaluate the benefits of a specific air regulation within
an economy-wide model.
5
This paper aims to describe the main approaches used by EPA and the academic literature to estimate the
economy-wide effects of improved air quality, highlighting potential areas of improvement and technical
challenges that exist. The methodological approaches discussed could have implications for both
adequately estimating the net economy-wide impacts and the benefits of air regulations. Before
considering the incorporation of environmental quality into applied economy-wide modeling, this paper
begins by providing background on the quantified and unquantified benefits of improved air quality. This
is followed by a discussion of the methods EPA typically uses to monetize these changes in benefits of air
regulations as a point of comparison for the economy-wide approach. The paper then reviews the studies
conducted by EPA and outside researchers that have incorporated, at least partially, environmental
quality into computable general equilibrium (CGE) models to estimate economic and welfare impacts. This
work is then examined in the context of the types of benefits typically associated with improvements in
air quality and the theoretical and empirical literature on benefits to assess key challenges and potential
areas of improvement. The paper concludes with a discussion of the importance of capturing spatial
heterogeneity when estimating the benefits of changes in air quality and how that may affect the
incorporation of benefits into economy-wide assessments.
2 Air Quality Benefits
To provide context for the later discussions in this paper regarding the incorporation of environmental
quality into economy-wide modeling, this section provides a basic overview of the beneficial impacts of
improved air quality and the methods that EPA typically uses to estimate the social benefits of
improvements in air quality for a regulatory analysis. It begins by summarizing the general categories of
benefits associated with improving air quality, including effects that are typically quantified in
contemporary benefit cost analyses (BCAs) and those that are not quantified but potentially could be if
improved data and methods are developed in the future. The section then provides a general overview of
the typical valuation framework that EPA employs, the methods it uses in performing benefits analysis,
and an example illustrating how results are often presented, including the characterization of
uncertainties.
2.1 Air Quality Benefits Traditionally Included in Benefit-Cost Analyses
Improvements in air quality may affect a wide variety of endpoints, including those related to human
health and others such as visibility, materials damage, and agricultural productivity. In recent EPA
regulatory analyses (US EPA, 2012) and its analysis of the benefits and costs of the Clean Air Act from 1990
- 2020 (US EPA, 2011), EPA typically quantifies and monetizes a number of health effects, listed in Table
1. Table 1 also lists health effects that, while potential benefits of air regulations, are not yet quantified in
EPA regulatory analyses. Table 2 provides an analogous table of quantified/monetized and unquantified
non-health benefits.
Some categories of human health benefits, such as changes in the incidence of adult or infant mortality
or of heart attacks, are more easily quantified and monetized due to the availability of scientifically
defensible concentration-response functions relating health outcomes to ambient pollution
concentrations and estimates of the monetized value for marginal changes in health outcomes. The
6
effects of other health outcomes, for instance changes in labor productivity due to a change in health
status, are more difficult to quantify. Depending on the endpoint, limited assessments may be available
for some subsets of affected populations (e.g., outdoor fruit pickers).
The monetized benefits of non-health impacts are typically more difficult to quantify due to several factors
including: 1) lack of available concentration- or deposition-response functions; 2) complexity of ecosystem
responses to pollutants; 3) heterogeneity in responses and dependence of responses to numerous
geographically varying factors; 4) lack of functions to translate changes in ecological endpoints into
changes in ecosystem services; and 5) lack of willingness-to-pay or other economic valuation estimates
for many types of ecosystem services. In past regulatory analyses, EPA has estimated the willingness to
pay for changes in visibility impairment and materials damages from reductions in PM2.5 and foliar damage
related to NOx emissions. In some cases, such as acidification of streams, EPA has been able to quantify
changes in stream chemistry and certain measures of fish health, but has not been able to monetize those
changes.
2.2 General Framework for Benefits Valuation
EPA typically follows a “damage-function” or “effect-by-effect” approach to estimate total benefits of
modeled changes in environmental quality when analyzing the effects of an air regulation (see for example
Levy et al., 2009; Fann et al., 2012; and Tagaris et al., 2009).1 In this approach, changes in individual
endpoints (specific effects that can be associated with changes in air quality) are estimated and dollar
values reflective of households’ willingness to pay (WTP) are assigned to those changes. Total benefits are
then calculated as the sum of the benefit estimates for all non-overlapping endpoints.
In some cases, the changes in environmental quality that are anticipated to result from an air regulation
are associated with endpoints that are valued directly by individuals, such as visibility in scenic national
parks. For such changes, it is relatively straightforward, at least conceptually, to link the pollution
externality to human welfare.
In other cases, additional analysis must first be conducted to map changes in air quality to changes in
endpoints more directly associated with human welfare. For example, EPA estimates how changes in
ozone and particulate matter (PM) concentrations affect health conditions that people value directly, with
health impact functions based on effect estimates in the epidemiology literature. Specifically, a health
impact function measures the change in a particular health endpoint (e.g., hospital admissions, mortality
rates) for a given change in air quality. A standard health impact function has four components: 1) an
effect estimate or risk coefficient from a particular epidemiological study; 2) a baseline incidence rate for
the health effect (obtained from either the epidemiology study or a source of public health statistics such
as the Centers for Disease Control); 3) the affected population; and 4) the estimated change in the relevant
air quality metric (e.g. annual mean PM2.5).
1 The damage function approach may provide a more bottom-up method of estimating total benefits in some cases
than other methods such as a hedonic price approach applied to housing prices, which would require strong
assumptions, such as perfect information, to fully characterize willingness to pay estimates.
7
Table 1: Human Health Effects of Air Pollutants Regulated by EPA
Benefits Category Specific Effect Effect Has Been
Quantified
Effect Has Been
Monetized
Reduced incidence of premature mortality and morbidity from exposure to PM2.5
Adult premature mortality based on cohort study estimates and expert elicitation estimates (age >25 or age >30)
Infant mortality (age <1)
Non-fatal heart attacks (age > 18)
Hospital admissions—respiratory (all ages)
Hospital admissions—cardiovascular (age >20)
Emergency department visits for asthma (all ages)
Acute bronchitis (age 8–12)
Lower respiratory symptoms (age 7–14)
Upper respiratory symptoms (asthmatics age 9–11)
Asthma exacerbation (asthmatics age 6–18)
Lost work days (age 18–65)
Minor restricted-activity days (age 18–65)
Chronic Bronchitis (age >26) — —
Emergency department visits for cardiovascular effects (all ages)
— —
Strokes and cerebrovascular disease (age 50–79)
— —
Other cardiovascular effects (e.g., other ages) — —
Other respiratory effects (e.g., pulmonary function, non-asthma ER visits, non-bronchitis chronic diseases, other ages and populations)
— —
Reproductive and developmental effects (e.g., low birth weight, pre-term births, etc.)
— —
Cancer, mutagenicity, and genotoxicity effects
— —
8
Table 1: Human Health Effects of Air Pollutants [continued]
Benefits Category Specific Effect Effect Has Been
Quantified
Effect Has Been
Monetized
Reduced incidence of mortality and morbidity from exposure to ozone
Premature mortality based on short-term study estimates (all ages)
Premature mortality based on long-term study estimates (age 30–99)
Hospital admissions—respiratory causes (age > 65)
Hospital admissions—respiratory causes (age <2)
Emergency department visits for asthma (all ages)
Minor restricted-activity days (age 18–65)
School absence days (age 5–17)
Decreased outdoor worker productivity (age 18–65)
— —
Other respiratory effects (e.g., premature aging of lungs)
— —
Cardiovascular and nervous system effects — —
Reproductive and developmental effects — —
Reduced incidence of mortality and morbidity from exposure to NO2
Asthma hospital admissions (all ages) — —
Chronic lung disease hospital admissions (age > 65)
— —
Respiratory emergency department visits (all ages)
— —
Asthma exacerbation (asthmatics age 4–18) — —
Acute respiratory symptoms (age 7–14) — —
Premature mortality — —
Other respiratory effects (e.g., airway hyperresponsiveness and inflammation, lung function, other ages and populations)
— —
9
Table 1: Human Health Effects of Air Pollutants [continued]
Benefits Category Specific Effect Effect Has Been
Quantified
Effect Has Been
Monetized
Reduced incidence of mortality and morbidity from exposure to SO2
Respiratory hospital admissions (age > 65)
Asthma emergency department visits (all ages)
Asthma exacerbation (asthmatics age 4–12)
Acute respiratory symptoms (age 7–14)
Premature mortality — —
Other respiratory effects (e.g., airway hyperresponsiveness and inflammation, lung function, other ages and populations)
— —
Reduced incidence of morbidity from exposure to methylmercury
Neurologic effects—IQ loss — —
Other neurologic effects (e.g., developmental delays, memory, behavior)
— —
Cardiovascular effects — —
Genotoxic, immunologic, and other toxic effects
— —
Reduced incidence of morbidity from exposure to lead
Neurologic effects—IQ loss
Increased diastolic blood pressure and sequelae
Other neurobehavioral and physiological effects
Delinquent and anti-social behavior
IQ loss effects on compensatory education
Hypertension
Non-fatal coronary heart disease
Non-fatal strokes
Premature mortality
Other cardiovascular diseases
Neurobehavioral function
Renal effects
Reproductive effects
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
10
Table 2: Non-Health Effects of Air Pollutants
Benefits Category Specific Effect
Effect Has
Been
Quantified
Effect Has
Been
Monetized
PM2.5 atmospheric
effects
Visibility impairment (includes effects from
NOx and SOx on nitrate and sulfate PM)
PM2.5 atmospheric and
deposition effects
Materials damage
Climate — —
Ecosystem effects (organics and metals) — —
NOx atmospheric effects Vegetation injury (through ozone)
NOx atmospheric and
deposition effects
Materials damage — —
Climate — —
NOx deposition effects
(includes effects of
deposition of particulate
nitrate)
Acidification (terrestrial and aquatic) plus
associated ecosystem effects —
Nitrogen enrichment (terrestrial and
aquatic) plus associated ecosystem effects — —
Secondary health effects (e.g. blue baby
syndrome) — —
SOx atmospheric effects Vegetation injury — —
SOx atmospheric and
deposition effects
Materials damage — —
Climate — —
SOx deposition effects
(includes effects of
deposition of particulate
sulfate)
Acidification (terrestrial and aquatic) plus
associated ecosystem effects —
Mercury methylation — —
Mercury atmospheric
and deposition effects
Ecosystem effects — —
Mercury methylation and associated effects
11
A typical health impact function has the form:
0 1xy y e , (1)
where y is the change in the health endpoint, 0y is the baseline incidence, equal to the baseline
incidence rate times the potentially affected population, is the effect estimate, and x is the
estimated change in the air quality metric. Other functional forms are sometimes applied, but the basic
elements remain the same. Upon mapping changes in emissions and the resulting air quality changes into
endpoints relevant for human welfare, economic valuation methods are used to estimate people’s WTP
for such changes.
Reductions in ambient concentrations of air pollution generally lower the risk of future adverse health
effects by a small amount for a large population. Epidemiological studies generally provide estimates of
the relative risks of a particular health effect for a given increment of air pollution (e.g., often per 10
μg/m3 for PM2.5). These relative risks can be used to develop health impact functions that relate a unit
reduction in air emissions to changes in the incidence of a health effect given the baseline prevalence. To
value these changes in incidence, WTP for changes in risk need to be converted into WTP per statistical
incidence. This measure is calculated by dividing individual WTP for a risk reduction by the related
observed change in risk. For example, suppose a measure is able to reduce the risk of premature mortality
from 2 in 10,000 to 1 in 10,000 (a reduction of 1 in 10,000). If individual WTP for this risk reduction is $100,
then the WTP for an avoided statistical premature mortality amounts to $1 million ($100/0.0001 change
in risk). Using this approach, the size of the affected population is automatically taken into account by the
number of incidences forecast by the health impact functions, which are applying results from
epidemiological studies to the relevant population. The same type of calculation can also produce values
for statistical incidences of other health endpoints.
For some health effects, such as hospital admissions, WTP estimates are generally not available. In these
cases, EPA has used the cost of treating or mitigating the effect as a proxy. For example, when assessing
the benefits of reduced hospital admissions EPA has used the avoided medical costs to monetize changes.
This approach, sometimes referred to as cost-of-illness (COI) estimation, generally (though not necessarily
in every case) understates the true value of reductions in risk of a health effect (Harrington and Portney,
1987; Berger, 1987). This is because it reflects the direct expenditures related to treatment ex post but
not the value of avoided pain and suffering associated with the health endpoint, and therefore, not the
full willingness to pay to avoid the illness ex ante.
In most cases, EPA relies on a benefit transfer approach to monetize health and non-health endpoints,
instead of conducting an original revealed or stated preference study of its own. Benefits transfer takes
primary research from the published literature (i.e. values or functional forms) that is deemed relevant
and similar enough to the regulatory context being evaluated and adapts it to the policy being analyzed.
For instance, when necessary, adjustments are made to account for the level of environmental quality
change, the socio-demographic and economic characteristics of the affected population, and other factors
12
to improve the accuracy and robustness of benefit transfer estimates. See EPA’s Guidelines for Preparing
Economic Analyses for more information on the particulars of benefits transfer.
2.2.1 Temporal Dynamics of Health Impacts
In assessing the benefits of air pollution regulations, reductions in premature mortality associated with
chronic PM2.5 exposure present unique challenges due to the importance of the temporal dynamics of
exposure and health impacts. For premature mortality, there is a “cessation” lag between reduced PM2.5
exposures and the total realization of changes in mortality risk. Although the structure of the lag is
uncertain, EPA follows the advice of the Health Effects Subcommittee of the Advisory Council on Clean Air
Compliance Analysis and assumes a segmented lag structure characterized by 30% of mortality risk
reductions in the first year, 50% over years 2 to 5, and 20% over the years 6 to 20 after the reduction in
PM2.5 (US EPA, 2004a). To take this into account in the valuation of premature mortality reductions, the
value in future years is discounted using rates of 3% and 7% based on OMB guidance (OMB, 2003).
Discounting is also applied in the valuation stage for endpoints such as nonfatal heart attacks, where a
portion of the costs associated with the endpoint may occur in years following the event.
2.2.2 Income adjustments
Economic theory suggests that the WTP for most goods will increase with real income (i.e., they are
normal goods). There is substantial empirical evidence that the income elasticity of the WTP for marginal
health risk reductions is positive, so as real income increases the WTP for environmental improvements
that lead to a reduced negative health outcome also increases. However, there is uncertainty about the
exact value of the income elasticity estimate. Although many analyses assume that the income elasticity
of WTP is unit elastic (i.e., a 10% higher real income level implies a 10% higher WTP to reduce risk
changes), some empirical evidence suggests that income elasticity may be less than one (i.e., inelastic) or
greater than one (i.e. elastic).2 If it is inelastic (elastic), this implies that, as real income rises the WTP value
also rises but at a slower (faster) rate than real income.
In addition, there may be differences in income elasticities across the WTP for different health endpoints,
for example, reflecting differences between acute and chronic conditions. Reported income elasticities
show that WTP is more elastic for severe health conditions and mortality risk, and more inelastic for minor
conditions (Kleckner and Neumann, 1999). As such, EPA uses different elasticity estimates to adjust the
WTP for minor health effects, severe and chronic health effects, and premature mortality. In recent
regulatory analyses, EPA has used elasticities ranging from 0.14 to 0.45 depending on the health endpoint.
2.3 Conducting and Presenting the Analysis
In a typical benefits analysis of an air quality regulation, health impacts are calculated at a relatively fine
spatial scale (e.g. at a 12km2 grid resolution) and then aggregated to state or national totals for
2 EPA’s Environmental Economics Advisory Committee of EPA’s Science Advisory Board noted that estimates of the
income elasticity of the VSL range from 0.1 to 1.0 in cross-section analysis of stated preference survey results and
between 1.3 and 3.0 in longitudinal studies of hedonic wage results (US EPA, 2011 p 21).
13
Table 3: Monetized PM2.5 Mortality Risk Reduction Benefits for the Revised and Alternative Annual
Primary PM2.5 Standards (Incremental to Analytical Baseline) (Millions of 2006$, 3% discount rate)
Health Effect
Revised and Annual Standards
(95th percentile confidence interval)
13 µg/m3 12 µg/m3 11 µg/m3
Krewski et al. (2009)
(adult mortality age 30+)
$1,300 $4,000 $13,000
($120--$3,500) ($370--$11,000) ($1,200--$35,000)
Lepeule et al. (2012)
(adult mortality age 25+)
$2,900 $9,000 $29,000
($250--$8,100) ($800--$26,000) ($2,600--$82,000)
Woodruff et al. (1997)
(infant mortality)
$3.4 $11 $35
($0.29--$10) ($0.91--$32) ($3.0--$100)
Notes:
1. Reproduced from Table 5-19 of the Regulatory Impact Analysis for the 2012 PM NAAQS.
2. All estimates are rounded to two significant digits. Estimates do not include unquantified health benefits noted in Table 5-2 or Section 5.6.5 or
welfare co-benefits noted in Chapter 6. These estimates reflect incremental emissions reductions from an analytical baseline that gives “an
adjustment” to the San Joaquin and South Coast areas in California for NOx emissions reductions expected to occur between 2020 and 2025,
when those areas are expected to demonstrate attainment with the revised standards. Full benefits of the revised standards in those two areas
will not be realized until 2025.
3. The reduction in premature fatalities each year accounts for over 98% of total monetized benefits in this analysis. Mortality risk valuation
assumes discounting over the SAB-recommended 20-year segmented lag structure.
presentation. Central estimates of health impacts are presented along with estimated 5th and 95th
percentiles, based on standard errors around the effect estimates from the epidemiological literature.
Central estimates of monetized values of health impacts are also presented along with confidence
intervals that reflect uncertainties in both the health impacts and unit values (e.g., WTP per case of disease
or hospital admissions). Table Table 3 provides an example of how economic benefits are presented based
on the recent RIA for the 2012 PM NAAQS rule (U.S. EPA, 2012).
In addition to the limited uncertainty analysis focused on propagating standard errors from the
epidemiological studies, benefits analyses typically include a number of sensitivity analyses for important
analytical choices, such as alternative effect estimates, functional forms (e.g. threshold models), discount
rates, lag structures, and inclusion of additional endpoints. While these sensitivity analyses are not formal
uncertainty assessments, they provide insights into how sensitive overall benefit estimates are to
different assumptions. In general, because of the large impact of PM2.5 on premature mortality and the
large magnitude of the VSL, total benefits are most sensitive to assumptions regarding the PM2.5 health
impact and valuation functions.
14
Because of the resource and time requirements required to generate air quality modeling results (e.g.,
months for a single year’s baseline and policy scenario), typical EPA benefits analyses choose one or two
analytical “snapshot” years, usually reflecting a year when full implementation of the regulation is
expected to have occurred (or as close to full implementation as possible given future year air quality
modeling constraints). Benefits are calculated only for these snapshot years, and thus are not
representative of an annualized stream of benefits or a net present value of benefits. They are simply the
annual benefits resulting from emissions reductions occurring during that snapshot year. In a few cases,
such as the analysis of the non-road diesel regulations issued in 2004, EPA estimated benefits in multiple
years and interpolated benefits to generate a stream of benefits over the full implementation period for
the regulation (U.S. EPA, 2004b).
3 The Benefits of Air Quality Improvements in CGE Modeling
While the partial analysis described in the previous section remains the most common approach for
applied benefit cost analysis, some studies have incorporated limited beneficial impacts of environmental
quality into CGE models for economy-wide estimates of the impacts. This section provides an overview of
those studies, including how benefits were included in the CGE modeling for the analysis of the benefits
and costs of the Clean Air Act from 1990 - 2020 (EPA, 2011) and studies by other researchers that have
incorporated benefits into CGE models.
3.1 EPA CGE Benefits Modeling – Benefits and Costs of the Clean Air Act from 1990 to 2020
In March 2011, EPA released a study evaluating the benefits and costs from 1990 to 2020 of programs
implemented under the 1990 Clean Air Act Amendments (CAAA), relative to a hypothetical baseline in
which control programs under the 1970 Clean Air Act and 1977 Amendments remained constant at their
1990 levels of scope and stringency. In addition to estimating partial equilibrium measures of the
prospective (2010 to 2020) direct compliance costs and household benefits, the study also included a CGE
analysis that incorporated both compliance costs and some benefits over this time period.
The CGE analysis performed for the study was the first analysis by EPA to include some categories of air
quality related benefits in a CGE model (specifically, the dynamic version of the Economic Model for Policy
Analysis Computable General Equilibrium or EMPAX-CGE model).3 The intent was to acknowledge that
while there have been general equilibrium costs associated with the implementation of the CAAA, there
have been concomitant improvements in health that may also have feedback effects on the economy.
3 EMPAX-CGE is a multi-industry, multi-region computable general equilibrium model of the U.S. economy. The
version of EMPAX-CGE used for the 2011 EPA study models 35 sectors and 4 representative households for each of
the 5 regions, and is resolved with 5-year time steps. It is assumed that households have perfect foresight of future
changes in policy and maximize utility over the full time horizon of the model by adjusting their consumption
patterns and their decisions about labor force participation. EMPAX-CGE is a full-employment model, so households
choose between labor (within their region’s single labor market) and leisure based on income and substitution
effects alone. Detailed information about the structure and calibration of the EMPAX-CGE model is available at
http://www.epa.gov/ttnecas1/EMPAXCGE.htm.
15
The benefits included in EMPAX-CGE represented 1) expected changes in medical expenditures associated
with pollution-related illness and 2) the expected change in workers’ time endowment stemming from
changes in mortality risks and changes in workdays lost due to illness.
As previously noted, EPA benefits analyses using the effect-by-effect approach described above are
typically conducted for one or two “snapshot” years once full implementation is expected to have
occurred. This presents a challenge for the inclusion of such benefits estimates in a dynamic CGE model,
since more than one or two years are modeled and the timing of impacts is important. In the case of the
2011 EPA study, multiple years of air quality modeling with and without the CAAA were available (2000,
2010, and 2020), and it was therefore possible to approximate the stream of benefits needed for inclusion
in EMPAX-CGE. The following sections briefly describe how these benefits were calculated and
incorporated into the model.4
3.1.1 Changes in Medical Expenditures
Estimates of the change in incidence or prevalence of several morbidity endpoints in 2010 and 2020 were
estimated using the environmental Benefits Mapping and Analysis Program (BenMAP).5 These estimates
of changes in incidence and prevalence were then multiplied by COI estimates from the published
literature to calculate total changes in medical expenditures associated with PM and ozone-related health
endpoints. Per event values of these annual medical expenditure estimates represented in Table 4. The
total change in medical expenditures were then incorporated into EMPAX-CGE as a change in the
representative household’s expenditure pattern. To interpolate the additional year 2015 needed for
inclusion in the EMPAX-CGE model, an estimated emissions trajectory was combined with estimates of
the benefit per ton of avoided emissions to create a benefits trajectory.6 An index was created using this
benefits trajectory, and this index was used to deflate the 2020 benefits estimates to their corresponding
values in 2015.
3.1.2 Effects on the Labor Force
The effect of PM2.5 related premature mortality and ozone and PM2.5 related morbidity on the labor force
is implemented in the EMPAX-CGE model as a reduction in the representative household’s time
endowment. The expected effect of the CAAA on population due to changes in PM2.5 related premature
mortality was modeled using a dynamic population simulation model (PopSim) that incorporates life table
4 A more complete discussion can be found in Chapter 8 of EPA (2011).
5 The Environmental Benefits Mapping and Analysis Program (BenMAP) is a software package containing a library of
concentration–response relationships, population data, baseline incidence rates, and valuation functions. It
quantifies and values the PM2.5 and ozone health impacts from air quality scenarios. Documentation for the most
current version of the software is available at http://www2.epa.gov/sites/production/files/2015-
04/documents/benmap-ce_user_manual_march_2015.pdf.
6 Benefit per ton ($/ton) estimates are derived by dividing the total monetized human health benefits associated
with an air quality scenario by the emission reductions associated with that scenario.
16
Table 4: Annual Medical Expenditures per Case, By Morbidity Endpoint [2006$]
2010 2020
PM2
Acute Myocardial Infarction3 $17,600 $17,300
Chronic Bronchitis4 $715 $810
Emergency Room Visits, Respiratory5 $369
$27,400
$21,000
Hospital Admissions, Cardiovascular6 $27,400
Hospital Admissions, Respiratory6 $21,000
Ozone7
Emergency Room Visits, Respiratory5 $369
Hospital Admissions, Respiratory6 $16,400 $17,100
Notes:
1. Except for chronic bronchitis and acute myocardial infarction, medical expenditures per case are applied to the change in
annual incidence. For chronic bronchitis and acute myocardial infarction, medical expenditures per case are applied to the
annual changes in the prevalence of each disease, to generate an annual rather than lifetime estimate of costs for these
chronic diseases.
2. Medical expenditure estimates for the following PM morbidity endpoints were not readily available: acute bronchitis, acute
respiratory symptoms, asthma exacerbation, lower respiratory symptoms, upper respiratory symptoms, and work loss days.
3. Derived from Wittels et al. (1990) and Russell et al. (1998), both as cited in Abt Associates (2008).
4. Cropper and Krupnick (1999).
5. We assume that each E.R. visit equals one day of lost work time per worker affected. The estimate of 0.2 days per case
reflects the percentage of cases realized by the working-age population, the ratio of workdays to total days in a year
(235/365), and the percent of the working-age population in the labor force.
6. Agency for Healthcare Research and Quality (2000), as cited in Abt Associates (2008).
7. Medical expenditure estimates for the following ozone morbidity endpoints were not readily available: minor restricted
activity days, school loss days, and outdoor worker productivity.
Disease Control. For a given policy scenario, PopSim simulates the U.S. population by age group and
gender for each year through 2050. This approach is different from the population projection
methodology employed by BenMAP, which uses exogenous growth factors to project 2010 Census data
to the appropriate analysis year. Instead, PopSim accounts for the effect that death rates in one year have
on the population at risk in future years, essentially following the population through time while
accounting for air-pollution related changes in population each year. As the EMPAX-CGE model is dynamic
and tracks changes in the economy over time, the population estimates produced by PopSim are more
appropriate for this exercise than the population estimates based upon static population growth factors
generated by BenMAP. Population was simulated for the with-CAAA scenario and without-CAAA scenario,
so the difference in any given year represents the change in population in that year due to the CAAA.
17
Because EMPAX-CGE only models working age individuals, the impact of changes in morbidity and
premature mortality incorporated into the model is limited to the cases expected to affect that
demographic. To this end, the simulated change in population was combined with age- and gender-
specific labor force participation rates from the Bureau of Labor Statistics. The percent change in labor
force was then calculated by dividing total labor force changes by the baseline (with-CAAA) projection of
labor force participation. That is, any additional population due to the CAAA will make on average the
same labor force participation choices as the population in the baseline scenario, leading to an overall
increase in the total time endowment for households in the economy. This percent change is assumed to
apply to the full time endowment (labor and leisure) for the labor force. Because of the full employment
nature of the model, households may choose to allocate a greater share of their time endowment to
leisure (and less to labor) in some scenarios. Since households derive utility from leisure as well as
consumption, this can lead to an increase in social welfare, even in the face of a decline in GDP and
consumption.
Morbidity impacts were also incorporated into EMPAX-CGE as a percent change in the labor and leisure
time available to workers. These impacts were calculated by health endpoint as work days lost using the
values in Table 5, which were converted to lost work years, and then expressed as a percent change in the
labor force by dividing estimated work years lost by the projected size of the labor force in each target
year.
3.1.3 Potential Analytical Limitations
In the prospective study on the costs and benefits of the Clean Air Act from 1990 to 2020 (US EPA, 2011),
EPA acknowledged several limitations associated with its incorporation of some benefits into the CGE
framework. The study did not include the effects of premature mortality or morbidity for individuals
outside the traditional workforce (e.g., children, elderly, informal economy). This notable exclusion biases
the estimated welfare changes downwards, particularly since children and elderly populations are known
to be more sensitive to health effects associated with changes in air pollution, though even without these
effects the net welfare change was found to be positive.
The full-employment nature of EMPAX-CGE, combined with the assumption that additional population
due to the CAAA make labor force participation decisions in the same way as existing households, may be
an overly restrictive assumption in some circumstances, though the overall impact on results is unclear. It
is also unclear how additional years of life due to improved air quality would actually be considered by
workers, though the model imposes the restriction that they continue to make choices as they always
have. Specifically, if the effect of the CAAA is mainly to reduce mortality rates at advanced (post-
retirement) ages, then the effects on labor force participation may be modest.
In addition, non-market and some market benefits, such as visibility improvements, productivity
enhancements in agriculture and forestry sectors, reduced materials damage, and reduced pain and
suffering from pollution-related illness were not included as inputs to EMPAX-CGE. These types of market
and non-market benefits, and how they may be included in CGE analyses, are discussed in Section 4 below.
18
Table 5: Work Days Lost Per Case, By Morbidity Endpoint
PM2.52
Acute Myocardial Infarction3 Age <25: N/A Age 45-54: 23.7 days
Age 25-34: 17.7 days Age 55-65: 137.0 days
Age 35-44: 14.5 days Age>65: 0 days
Chronic Bronchitis3 Age <25: N/A Age 45-54: 55.5 days
Age 25-34: 50.3 days Age 55-65: 73.5 days
Age 35-44: 42.2 days Age >65: 0 days
Hospital Admissions, Cardiovascular4 Age 0-14: N/A Age 45-64: 17.9 days
Age 15-44: 18.3 days Age >64: 7.0 days Hospital Admissions, Respiratory4 Age 0-14: N/A Age 45-64: 30.1 days
Age 15-44: 30.7 days Age >64: 7.5 days Emergency Room Visits, Respiratory5 Average across all age groups: 0.2 days
Work Loss Days Average among working age population: 1 day
Ozone6
School Loss Days7 Average across all age groups: 0.7 days
Worker Productivity Not applicable8
Hospital Admissions, Respiratory9,10 Age <2: 0 days
Age >64: 7.5 days Emergency Room Visits, Respiratory5 Average across all age groups: 0.2 days
Notes:
N/A indicates that the underlying C-R function does not provide incidence estimates for that age group.
1. Except for chronic bronchitis and acute myocardial infarction, the number of work days lost is applied to the change in annual incidence.
For chronic bronchitis and acute myocardial infarction, the work days lost presented in this table are applied to annual changes in the
prevalence of each disease.
2. Separate work loss day estimates were not generated for the following PM health endpoints: acute bronchitis, acute respiratory
symptoms, asthma exacerbation, lower respiratory symptoms, and upper respiratory symptoms. The lost work days associated with these
endpoints are already reflected in the work loss day endpoint included in this table.
3. Derived from Cropper and Krupnick (1999).
4. Agency for Healthcare Research and Quality (2000), as cited in BenMAP user’s guide, Abt Associates (2008).
5. We assume that each E.R. visit equals one day of lost work time per worker affected. The estimate of 0.2 days per case reflects the
percentage of cases realized by the working-age population, the ratio of workdays to total days in a year (235/365), and the percent of the
working-age population in the labor force.
6. We did not estimate the number of work days lost per case of acute respiratory symptoms associated with ozone exposure.
7. Derived from Abt Associates (2008). Note that 0.7 is the estimated average work loss days per school loss day, incorporating work-force
participation rates for caregivers.
8. The benefits analysis presented in Chapter 5 does not estimate the number of cases for the worker productivity endpoint. Instead, worker
productivity is estimated as the change in income associated with changes in ozone concentrations. We estimated the work days lost per
dollar of income lost based on the average daily wages of outdoor workers.
9. Derived from Abt Associates (2008).
10. The dose-response function for ozone-related respiratory hospital admissions does not cover populations older than two years old and
younger than 65. For this endpoint we do not address potential caregiver time lost for incidence in either cohort.
19
Lastly, environmental quality (outside of the limited health impacts considered) was essentially assumed
to be a separable component of the utility function for households. This may not hold, however, as cleaner
air may encourage leisure activities, making air quality a complement to leisure.
3.2 Outside Studies Modeling Direct Health Impacts of Air Pollution in CGE Models
Several researchers have used relatively similar approaches as that of EPA (2011) to model the health
benefits of improvements in environmental quality in CGE models, though with some interesting
variations and for different applications. Vennemo (1997) was one of the first studies to incorporate
environmental externalities into an applied CGE model, the Dynamic Resource/Environment Applied
Model (DREAM) of the Norwegian economy. The model was based on a small open economy producing a
single good available for export and domestic consumption and five input factors. The model estimated
the impact of economic activities on emissions of criteria air pollutants using emissions coefficients and
then mapped emissions to health, non-market, and production impacts.7 Morbidity associated air
pollution were assumed to impact labor supply at both the extensive and intensive margins. In the model,
these two effects were combined and modeled as pollutant specific labor shocks. Changes in mortality
risk associated with air pollution along with non-market impacts (discussed in more detail in Section 3.4)
were modeled as a separable shock to household utility. Over the course of the model’s time horizon the
impact of the environmental feedbacks on consumption of goods and services increased from -1.4% in
2030 to -3.5% in 2090, with a welfare estimate of overall damages at 9.2% of full wealth (including leisure).
Mayeres and van Regemorter (2008) extended the GEM-E3 CGE model of the European economy to
include non-separable health impacts of air pollution. The version of the GEM-E3 model used for this study
was a dynamic model of 14 European countries with a 1995 benchmark. The model included an
environmental component that modeled the emissions, transport, and increase in atmospheric
concentrations of criteria air pollutants and carbon dioxide from the energy sector. In the standard version
of the model all mortality, morbidity, and non-health impacts of air pollution were treated as separable
in the utility function for the representative household. Mayeres and van Regemorter incorporated the
effects of exposure related morbidity and the medical costs of pollution-induced mortality more fully into
the model. Specifically, they considered the effect of air pollution on private and public spending for
medical services and the time endowment within a bi-directional coupling of the economy and the
environment.
These non-separable health impacts were included by extending both the household’s representation and
the economy’s production functions. To model the effect of air pollution on the household’s demand for
medical services the top nest of the Stone-Geary utility function, which previously included only
composite consumption, C , and leisure, l , was extended to include health, H . Specifically, the utility
function was defined as
7 Vennemo (1997) also considers the feedbacks from changes in traffic volumes, though the treatment is analogous
to that of air pollution.
20
1 2 3 ,1
ln ln lnM
H m mm
C C l l H H AU
, (2)
where bars represent subsistence levels and the last component represents separable effects of changes
in the ambient concentrations of pollutant from the benchmark, ,H mA , for the M pollutants, such that
,H m represents the marginal impact of changes in ambient concentration.8 Health was defined based on
a production function approach, such that
1, 21
*M
m mm
A MH DH E
, (3)
where *H is household health in the absence of air pollution or consumption of medical services, 1,m
represents the marginal impact of the changes in the ambient concentration of pollutant m on health,
and MED is the consumption of medical services. The price of medical services included government
subsidies based on the social security systems in place within the European countries modeled, such that
an increase in the quantity demanded of medical services affected both the household and government
budgets.
The time endowment effect of morbidity was modeled in two parts to recognize the existence of paid sick
leave requirements in the countries modeled. Part of the reduction in the time endowment was allocated
to reducing the households’ time endowment directly, while the remaining lost time endowment was
modeled as a labor productivity shock in the model’s production functions. The model was calibrated
based on the assumption that 58% of the lost time should manifest as a labor productivity shock in the
model.
The new parameters of the extended model associated with changes in air quality were calibrated to
match the results of studies that have estimated components of marginal WTP for changes in ambient
concentration separately (i.e., changes in household income, medical services expenditures, and the
separable component). The authors assumed that 10% of the marginal WTP estimate for mortality risk
reductions was associated with a change in expected medical expenditures and used that for calibration.
Mayeres and van Regemorter (2008) then compared the results from the extended GEM-E3 model to the
case where all air pollution damages were separable. Ex ante marginal WTP for reductions in ambient
concentrations were calibrated to be equivalent between the two models. The policy examined was a
country specific CO2 tax commensurate with commitments under the Kyoto Protocol. While climate
change impacts were not modeled, the results included the benefits from reduction in energy sector
criteria air pollutant emissions correlated with CO2 emissions. They found the macroeconomic effects
(e.g., GDP, labor) to be very similar between the two cases with the extended model estimating a larger
increase in consumption. Given the difference in the utility functions between the two models, a direct
comparison of estimates for the change in welfare was not possible.
8 While the GEM-E3 model is dynamic, for convenience we focus on the household problem for only a single period.
21
Li (2002) applied a similar technique to study the welfare impacts of an economy-wide CO2 emissions tax
in Thailand, accounting for general equilibrium impacts of correlated PM10 emission changes. The static
CGE model modified the standard International Food Policy Research Institute CGE framework outlined in
Lofgren et al. (2001) and had 61 production sectors and three households (agricultural, non-agricultural,
and government-employed). The model was extended to include a bi-directional coupling between
environmental quality (PM10 only) and the economy, though the combined model was only run for one
iteration so there may not be convergence in the model. Exposure-response functions were used to map
PM10 concentrations, derived from emissions, to premature mortality and morbidity endpoints for non-
agriculture and government-employed households of working age. In addition to affecting labor supply
on the extensive margin it was assumed that exposure also affects labor supply on the intensive margin
through a reduction in labor productivity. Premature mortality, morbidity, and labor productivity effects
were all represented through a reduction in the households’ time endowments. The model also included
pollution-induced medical expenditures in the same fashion as Mayeres and van Regemorter (2008),
where improvements in environmental quality reduced household and public expenditures on medical
services.
Other studies examining the direct health impacts of air pollution using a CGE model extended MIT’s
Emission Prediction and Policy Analysis (EPPA) model to include health impacts of air pollution.9 The
extended model has been referred to as the EPPA Health Effects (EPPA-HE) model (Matus et al., 2008). To
date the EPPA-HE framework has been used to study the impact of historical air pollution (ozone and
PM10) in the U.S. (Matus et al., 2008), Europe (Nam et al., 2010), and China (Matus et al., 2012), and future
ozone pollution globally (Selin et al., 2009). While there are some slight differences in terms of the data
employed and calibration of the model, the general methodology used to incorporate health impacts
associated with air pollution and the interaction with the economy in the model was the same across
these applications.
The EPPA-HE model incorporated the impacts associated with acute and chronic premature mortality, and
pollution-induced morbidity. For premature mortality, the model accounted for the damages to
households in terms of total factor consumption, including leisure, while also accounting for the
cumulative economic effect of reduced labor, consumption, and investment associated with premature
mortality. For pollution-induced morbidity, the model accounted for the reduced time available for labor
and/or leisure, the increased demand for medical services, and cumulative economic effects from those
changes.
To capture these health impacts the EPPA-HE model modified the original model structure in two ways.
First, leisure was introduced into the model by including an additional top-level nest in the household
utility function along with a constraint on the time endowment, which was split between labor and leisure.
Second, a household healthcare production sector was added to account for the time and medical service
expenditures required to respond to pollution-induced morbidity. The household healthcare sectors were
9 EPPA version 4 is a global recursive dynamic CGE model with a five-year time step and 16 regions built on the Global
Trade Analysis Project (GTAP) database.
22
Figure 1: Household and Consumption Structure for EPPA-HE (Source: Matus et al., 2008)
health endpoint and pollutant specific, and Leontief in labor and medical services. The cost shares were
calibrated based on external studies that estimate the composite costs of health endpoints. The updated
structure of the household consumption function is presented in Figure 1.
To date the studies using the EPPA-HE model have used a unidirectional coupling, where atmospheric
concentrations of pollutants are taken as exogenous and fixed, as opposed to the bi-directional coupling
of Li (2002) and Mayeres and van Regemorter (2008). The model used a series of exposure-response
functions from the epidemiological literature to map the atmospheric concentrations into health
endpoints. In all cases, it was assumed that the exposure-response functions are linear without a
threshold. For morbidity related endpoints, which were all assumed to be from acute exposure, this stage
provided an estimate of the number of cases/events expected to occur in each model year. For premature
mortality, the exposure-response functions provided an estimate of the proportional increase in mortality
rates in the model year, which must be applied to baseline mortality rates to estimate the expected level
of mortality. To account for the effects of premature mortality due to chronic exposure a cohort
population model was used to track the number of individuals in a given age group in each model year.
23
For each cohort the annual average exposure was computed and entered into the appropriate exposure-
response function to estimate the change in the mortality rate.10
The results of this mapping were then used to adjust the time endowment inputs to account for losses in
labor supply due to premature mortality. Care was taken to ensure that for premature mortality of, what
were assumed to be, working age individuals (younger than 65) it was effective units of labor that were
removed from the time endowment. Effective labor includes time that was adjusted for labor productivity
growth, which was included as an exogenous increase in the time endowment over model’s time horizon.
Based on an assessment of the literature, the model assumed that on average acute mortality occurs 0.5
years earlier than in the counterfactual due to exposure to air pollution. This is representative of the
characteristic that premature mortality due to acute exposure primarily affects vulnerable populations.
To account for the effects of chronic exposure-induced premature mortality, the model assumed an
average lifespan of 75 years. Morbidity-related health endpoints were applied effectively as Hicks neutral
shocks to the household healthcare sector to represent an increased need for time and medical services
to maintain a given level of health.
In a given simulation year, the modeled welfare loss associated with pollution damages was based on the
value of the lost time endowment in that period, the distortion imposed by the additional demand for
medical services, and the cumulative economic impacts associated with reduced consumption and
investment in previous years due to pollution. Lost time endowment in a given year was valued at the
wage rate on the margin. To account for the loss of time endowment from cohorts that are not of working
age (children and the elderly), the value of time was adjusted to be 1/3 and 2/3 of the wage rate,
respectively.
Matus et al. (2008) used the EPPA-HE model to study the impact of air pollution and the benefits of air
pollution regulations in the U.S. from 1970 to 2000.11 Specifically, they focused on the damages associated
with ozone, nitrogen dioxide, sulfur dioxide, carbon monoxide, and particulate matter (PM2.5 and PM10).
To estimate exposures, it was assumed that all urban populations are exposed to the average urban
pollution estimates published by the EPA. Conditional on historic exposure levels and their estimated
damages, the EPPA-HE model was calibrated to reproduce observed economic activity. With this
calibration, the authors estimated the benefits (reduced damages) associated with air pollution
regulations assuming a counterfactual level of pollution exposure and the remaining air pollution damages
based on above background pollution levels. Noting the uncertainty surrounding estimates in the
exposure-response functions, Matus et al. conducted sensitivity analysis by setting the estimated
10 To date chronic exposure induced premature mortality in the EPPA-HE model has been based on PM10. The
underlying epidemiological studies on the effects of chronic exposure to PM10 typically estimate the average
exposure-response function for the population over 30 years of age. The EPPA-HE model accounts for cohort specific
differences by weighting the exposure response function by the share of the cohorts’ historic mortality due to
cardiopulmonary and lung diseases relative to the total population over 30 and consistent with the epidemiological
literature assumes to increase in mortality rates for those under 30.
11 Matus et al. (2008) focus on incorporating the health impacts of air pollution mitigation policy in a CGE models
and do not explicitly consider the costs of the mitigation in their framework.
24
exposure-response coefficients at their 95% confidence intervals. (The forthcoming whitepaper on
uncertainty in general equilibrium modeling and benefits estimation will include a more detailed
discussion of uncertainty analysis in a CGE context.)
Nam et al. (2010) used the EPPA-HE model to study the impact of air pollution in Europe from 1970 to
2005 in a similar fashion. In contrast to Matus et al. (2008), this study focused only on ozone and PM10.
However, in this case gridded population and concentration data was combined to calculate population
weighted exposure estimates that are entered into the exposure-response functions. Nam et al. focused
on the residual damages from air pollution relative to background levels conditional on air regulations
already in place that affect the gridded concentration data. They found that the damages measured as a
percentage change in overall welfare or consumption, while still notable, declined over time likely due to
the strengthening of air pollution regulations.
Matus et al. (2012) use a similar approach as Matus et al. (2008) and Nam et al. (2010) to studied the
historic damages in China from ozone and PM10 air pollution. The application faced additional challenges
relative to the previous studies due to a paucity of concentration data from the pollutants being studied.
The cost shares for the household healthcare sectors were recalibrated specifically for China using
estimates of the Chinese composite morbidity costs. They considered a series of simulations using
historical emissions, background emissions, and two hypothetical policy scenarios. Similar to the previous
study the damages as a percentage of total welfare (or consumption) declined over time. However, in this
case the results were in part driven by rapid economic expansion. This rapid economic growth also led to
a case where losses in previous years account for 29% of the damages in 2005, as opposed to the 12%
found by Nam et al. for Europe.
Selin et al. (2009) depart from previous work using the EPPA-HE model to focus on future changes in air
pollution worldwide instead of within a single country or region. Specifically, they studied the damages
associated with future changes in ozone exposure out to 2050, both with and without climate change.
Ozone exposure was based on global atmospheric chemistry modeling with and without climate change
and was population weighted by region based on gridded population data from 2000 (i.e., no future
migration). The calibration of cost shares in the household healthcare sectors in developed countries was
based on the estimates used by Nam et al. but adjusted using purchasing power parity (PPP). For
developing countries, cost shares were calibrated based on estimates for China, again adjusted by the
PPP. Age distributions were also separately specified for developing and developed regions. While
previous studies based on EPPA-HE relied on sensitivity analysis to explore the robustness of their results,
Selin et al. took a probabilistic approach and defined distribution over the exposure-response function
estimates and cost shares in the household healthcare sectors, and then used a Monte Carlo simulation
to integrate over the distributions. Selin et al. found that, when compared to a scenario in which ozone
concentrations are assumed to be at background levels, 40% of the economic damages associated with
ozone pollution in 2050 were based on the cumulative effect of mortality and morbidity in previous years.
3.2.1 Potential Analytical Limitations
While the studies discussed in this subsection have been important for highlighting the potential
importance of the economy-wide effects of environmental quality improvements and providing
25
frameworks for their incorporation into CGE models, their remain a number of potential limitations. To
date the treatment of changes in premature mortality risk have been confined to changes in the time
endowment of a representative agent. Furthermore, distinctions across space (e.g., distinction between
emissions, concentrations, and population locations) and sub-populations (e.g., children and the elderly)
that can be highly relevant for estimating health impacts have been only partially addressed to date in
CGE models, if at all.
To date the approach taken by most, but not all, studies has been to assume that environmental quality
is exogenous to economic activity. Therefore, in the cases where studies found the inclusion of health
effects from improvements in environmental quality to have a significant impact on economic activity,
there is no feedback from those changes on environmental quality. It may be the case that such feedbacks
are of second order importance and the challenges associated with implementing endogenous is
environmental quality in a CGE model is not warranted.
In addition, in most cases non-market and some market benefits, such as visibility improvements,
productivity enhancements in agriculture and forestry sectors, reduced materials damage, and reduced
pain and suffering from pollution-related illness were not included in the analyses. These types of market
and non-market benefits may be significant for assessing the economy-wide effects of improvements in
environmental quality as discussed in the subsequent section. Additional discussion of potential
challenges and room for improvement is provided in Section 4.
3.3 Non-Health Environmental Feedbacks in CGE Models
A number of studies have incorporated non-health environment-economy linkages in CGE models. These
linkages have been found to be important for assessing not only the benefits of policies via their impact
on environmental quality, but also the costs of certain policies that may indirectly increase the
opportunity cost, or reduce the availability, of non-market goods and services.
Vennemo (1997), described in the previous section, was one of the first studies to include non-health
environmental feedbacks in an applied CGE analysis. Specifically, Vennemo considered the role of
environmental policies on depreciation rate of capital, both through reduced sulfuric deposition on
buildings and reduced miles traveled on roadways. The study also considered the impact of acidification
on forests by modeling the change as a productivity shock to the forestry sector and use of forests by
households for recreation.
Finnoff and Tschirhart (2008) studied the impact of fishery regulation on the economy by coupling
ecological and economic general equilibrium models. Specifically, they linked a recursive dynamic CGE
model of the Alaskan economy to a food web model of Alaska’s marine ecosystem to study the welfare
impacts in Alaska of regulating commercial fishing of a prominent species to protect the endangered
stellar sea lion’s food source. While the economic link is clear with respect to the directly regulated
species, the policy also affects other economic sectors as the sea lion, along with its predator the killer
whale, play a role in the state’s tourism sector.
Their CGE model represented three productions sectors covering fishery, recreation and tourism, and a
composite good. Fishing sector operations were constrained by a regulator that selects the total allowable
26
catch and season length for the industry. The regulator’s choices defined the equilibrium populations of
the various species included in the food web along with the economic output of the fishing sector. The
tourism sector depended on populations of some of these species, particularly “charismatic” species – sea
lions, killer whales, and sea otters – that were required to create the “recreation experience” sought by
consumers. The linear aggregate of charismatic mammal populations derived from the ecological model
was incorporated into a Cobb-Douglas production function for the tourism sector along with capital and
labor. In the production function for tourism, the exponent on charismatic populations defined the
importance of these species for the sector, and was specified exogenously. Other parameters were
calibrated to match benchmark data.
Finnoff and Tschirhart (2008) modeled the impact of various regulatory policies and found potentially
notable general equilibrium effects occurring from both the change in the directly regulated industry and
the tourism sector via its ecosystem connection. They also found that the ecosystem connections with the
economy played an important role in economic impacts of potential interest to policy makers. While most
of the primary factors (capital and labor) utilized in the directly regulated industry reside outside of Alaska,
they still found impacts in the domestic labor market, not only due to the real income effect, but also
through changes in the tourism sector due to ecosystem changes that result from the policy.
Antoine et al. (2008) directly introduced non-market recreation services (hunting, fishing, and wildlife
viewing) into the EPPA CGE model. While the goal of their study was to establish better estimates for the
cost of mitigating greenhouse gas emissions through biofuels policy by representing competing demands
for land, the methodological contributions of this work are useful to discuss in the current context. The
default category of unmanaged forest in the model was disaggregated into unmanaged but protected
forests (e.g., national parks, wildlife reserves) and unmanaged and unprotected forests. Only the latter
could be transformed to a managed land type (cropland, pastureland, managed forest, or unmanaged but
protected forests).12 The household sector was expanded to include an outdoor recreation sector, which
included hunting and fishing services, wildlife viewing on reserved land, and wildlife viewing on other land.
In each case production of the service was a CES combination of land and a Leontief composite of food,
transportation, services, leisure, and other expenditures. For wildlife viewing on other land and hunting
and fishing, the land input was unmanaged and unprotected forests, while for wildlife viewing on
protected land the land input was an improved unmanaged but protected forests, which was a Leontief
composite of unmanaged but protected forests and government expenditures.
Antoine et al. (2008) used the extended version of the EPPA model to consider the impact of greenhouse
gas mitigation policies on biofuels use and household welfare, both with and without the inclusion of
outdoor recreation in the model. They found that, both in the baseline and the policy case, the inclusion
of demand for unmanaged forests for outdoor recreation lowered the amount of land converted to
biofuel production. This result highlighted how the opportunity cost of biofuels as a mitigation measure
was contingent upon the competing demands for land including non-market services, such as outdoor
recreation. The authors found this tradeoff associated with non-market services to be a potentially
important component for characterizing the social cost of mitigation policy, increasing the welfare costs
12 Conversion to unmanaged but protected forests is considered permanent.
27
by around 20% in years where biofuels were considered a relatively cost-effective mitigation strategy by
the model.
As noted by Antoine et al. (2008), there are a number of challenges with this approach to incorporating
recreation in a CGE model. The CGE model requires benchmark estimates of the value for each of the
inputs to the outdoor recreation activities. Even in the U.S., national or regional data for the time spent
engaged in these activities is not complete, and for non-U.S. regions, data can be even sparser.13 There
are also data limitations in identifying the type of land used for the outdoor recreation activities. As an
approximation, Antoine et al. assumed that all of the outdoor recreation activities take place on forest
land, where in reality other land types (e.g., pastureland and grasslands) are also used for some of these
activities. The model also assumed that the forest land used for these activities is provided as if supplied
by a competitive producer responding to market demand. However, the conversion of land and ability of
the owner to effectively price its use might be more challenging in practice. Finally, the literature does not
provide readily available estimates for the substitution elasticities that relate the ability for households to
substitute between land and other market goods when providing the recreation services and their ability
to substitute between outdoor recreation and different activities in their utility function.
Carbone and Smith (2013) tested the quantitative importance of general equilibrium welfare effects from
improving environmental quality compared to partial equilibrium measures. In this study, the authors
extended the U.S. based CGE model of Goulder and Williams (2003) to include SO2 and NOx emissions and
their potential impacts on human health, non-market ecosystem services, and non-use existence services.
The base model is a static CGE model with five consumption goods, four intermediate goods, and one
basic factor of production (labor) calibrated to a 1995 benchmark year. Carbone and Smith added sector-
specific emissions factors for SO2 and NOx emissions that have effects on lakes and forests through acidic
deposition, along with negative human health impacts. They also included two non-market services
associated with recreational fisheries and tree cover and a non-use existence value associated with
general habitat services. The preferences of the representative agent were defined with a nested CES
utility function that combined the non-market services and output from the consumer services sector in
a nest that is then combined with leisure to capture the recreational nature of these services. This
aggregate leisure/recreation consumption good is then combined with an aggregate market consumption
good. Finally, general habitat services are incorporated into the top-level nest. A diagram of this
preference structure is provided in Figure 2.
The inclusion of these non-health environmental feedbacks presented a unique set of challenges in
calibrating the parameters of the CGE model. Basic deposition factors were used to convert emissions
estimates in tons to deposition rates in kg/ha on a national scale. The benchmark expenditure shares for
these goods in the nested CES utility function were then calibrated such that the marginal WTP for those
13 While the focus of the study was on greenhouse gas mitigation policies in the U.S., the authors noted that
incorporating the same demand for outdoor recreation services in other regions was necessary to ensure that food
and forestry imports were appropriately priced, such that the opportunity costs of preserving land within the U.S.
were appropriately characterized.
28
Figure 2: Carbone and Smith (2013) Preference Structure (Source: Carbone and Smith, 2013)
services in the model match estimates from the literature. The non-market and existence services were
assumed to be quasi-fixed goods, in that they were endogenous to the model’s equilibrium but were
considered exogenous by the household. Therefore, analytical expressions relating the benchmark
expenditure shares to the marginal WTP were not available, requiring a numerical procedure to calibrate
benchmark expenditure shares, along with the other parameters that ensure labor supply elasticities
implicit in the model match empirical estimates.
The results of the calibration procedure are conditional on assumptions regarding the substitution
elasticities in the model, for which there may not be a significant underlying empirical literature to draw
upon with respect to non-market services. Carbone and Smith (2013) assumed that general existence and
habitat services are just as substitutable for the aggregate consumption of market and non-market goods
as use-based non-market services (combined with consumer services and leisure) are for the bundled
consumption good. Furthermore, they assumed that for the nests combining consumer services and non-
market services and the combination of that aggregate bundle with leisure, that the cross-elasticities
29
within the nests are one-fourth as substitutable compared to how substitutable the aggregate is with the
market-based consumption bundle. In terms of Figure 2, these assumption imply that h u and
/ 4rl r u .
Changes in level/quality of the non-market services have an effect on the economy through changes in
agent’s consumption patterns and supply of labor. Changes in general habitat services, even as a non-use
category, have an effect on market services through the change in marginal utility. Carbone and Smith
(2013) conducted a series of experiments with their model to compare the welfare change with and
without the general equilibrium effects brought about through changes in environmental quality from
polices to reduce SO2 and NOx emissions. For a policy resulting in a 40% emissions reduction, roughly the
magnitude of the CAAA impact, and focusing only on improvements to recreational fisheries, they found
that including general equilibrium effects increases the estimate of the welfare improvement by 12% to
55%. The range is based on a sensitivity analysis around the substitutability between market and non-
market consumption. The lower end of the range assumes greater substitutability relative to the reference
case, while the higher end of the range assumes that they are stronger complements than in the reference
case.14 These results led the authors to conclude that even when the value of non-market services is a
small fraction of the aggregate value of the economy, the general equilibrium effects stemming from how
changes in environmental quality change household behavior in economic markets may be notable.
Carbone and Smith (2013) also included some health impacts of SO2 and NOx emissions in a manner
similar to those discussed in Sections 3.1 and 3.2. Specifically, they considered the lost time endowment
associated with hospital admissions due to respiratory ailments induced by SO2 and NOx emissions. They
found that the welfare effects are small and have a negligible effect on the general equilibrium effects of
changes in environmental quality, which is in line with the fact that hospital admissions represent a
relatively small portion of the health impacts associated with SO2 and NOx emissions.
3.4 Benefit-Side Tax Interaction Effects
The effects of environmental policy in a second-best setting that accounts for pre-existing tax distortions
has been widely studied. Initial work in this area focused on tax interaction and revenue recycling effects
on the cost side of the equation, finding general equilibrium effects to be of first order importance.15
Subsequently, researchers have examined the role of interactions between pre-existing market
distortions and the benefits of improved environmental quality. As noted by Bovenberg and Mooij (1994),
if environmental quality enters the utility function in a non-separable fashion, changes to environmental
quality will affect the household’s labor-leisure choice, thereby working to exacerbate or mitigate,
depending on the specifics of the situation, the effects of pre-existing factor market distortions.
14 Carbone and Smith (2013) recalibrated the model depending on whether the welfare measure includes general
equilibrium effects and for the different sets of substitution elasticities so that the marginal WTP implicit in the
model remained in line with empirical estimates.
15 See for example Bovenberg and de Mooij (1994), Bovenberg and van der Ploeg (1994), Goulder (1995), Goulder
et al. (1999), Parry (1994), Parry (1997), and Parry et al. (1999).
30
Williams (2002) provided an intuitive approach to understanding the potential interaction of
environmental improvements and distortionary factor market taxes in a static setting. The model
considered an economy producing two goods that are consumed by a representative household, and a
government that is financed through a labor tax. Production of one good negatively affects environmental
quality, which can affect household utility directly (weakly separable with respect to consumption and
leisure), the household’s time endowment, their demand for medical services, and the efficiency of
production. Williams showed that the welfare impact associated with internalizing the pollution
externality using a revenue neutral tax on the production of the dirty good can be separated into four
effects: the primary welfare effect, the revenue recycling effect, the cost-side tax interaction effect, and
the benefit-side tax interaction effect.
The primary welfare effect captures the net benefits of an improvement in environmental quality in the
absence of the distortionary labor tax, including any potential general equilibrium effects of the policy. In
the case where the government raises revenue from labor taxes, the remaining three effects may be non-
zero. The revenue recycling component accounts for the welfare gain associated with the government
using revenues from the pollution tax to reduce the distortionary labor tax rate, thereby incentivizing
labor.16 The cost-side tax interaction effect captures the fact that policies may implicitly lower the real
wage through higher prices on goods and services, which may further exacerbate the existing labor market
distortion through a decrease in labor.
The fourth effect, the benefit-side tax interaction effect, captures the interaction between the impacts of
improvements in environmental quality and pre-existing distortionary taxes. The magnitude and sign of
the benefit-side tax interaction effect is conditional on how changes in environmental quality affect firms
and households. For example, if an improvement in environmental quality increases the production
efficiency of firms, this reduces their marginal cost of production, effectively acting as an increase in the
household’s real wage, which incentivizes labor. In this case, the benefits of the improvement are higher,
as they work to mitigate the impacts of the pre-existing distortion. If instead the improvement in
environmental quality increases the productivity of a fixed factor (e.g., increase agriculture/land
productivity), this acts as an income effect for households, thereby exacerbating the pre-existing
distortion and reducing the benefits from the improvement.
The benefit-side tax interaction effect is ambiguous if improvements in environmental quality increase
the household’s time endowment. This effect is particularly relevant for policies that improve public
health through improvements in air quality, since as noted above, this effect may be viewed as increase
the amount of time available to the household. This change improves the well-being of the household
such that there is an income effect, which increases leisure (decreases labor). However, the household
will also have more time to devote to all activities, so the net effect on labor supply is ambiguous.
16 The role of this effect is unclear with respect to most EPA regulations, as they rarely generate revenue. However,
in some cases EPA regulations implicitly relax the government’s budget constraint (e.g., by preventing future publicly
funded remediation).
31
In his conceptual model, Williams (2002) assumed that the demand for medical services was a function of
environmental quality only and therefore only acts to reduce household income. Thus, an improvement
in environmental quality means that a household no longer needs to spend a portion of their income on
medical services to address the health effects of pollution. The resulting increase in leisure from the
income effect exacerbates the pre-existing distortion and reduces the overall benefits of the
environmental improvement.
It is unclear how this last result generalizes to applied settings where a portion of a country’s medical
expenditures are financed through labor taxes (e.g., Medicare, Medicaid). In their applied studies (which
take place outside of the United States), Li (2002) and Mayeres and van Regemorter (2008) took this into
account, but in neither case is the benefit-side tax interaction effect investigated. However, a number of
theoretical studies have considered the potential for interactions between the health benefits of
improved environmental quality and distortions in the healthcare market. Parry and Bento (2000) studied
the tax interaction effect in the presence of tax deductions for some consumption goods, and established
the intuition that the effect of environmental policies could mitigate such economic distortions if it led to
a reduction in the sub-optimal overconsumption of such goods due to the subsidy. Caffet (2005) and
Yamagami (2009) built of this work and explicitly incorporated subsidies for medical expenditures into the
theoretical model of William (2002). They found that in the case of distorted markets for healthcare goods
and services that it is possible for the health benefits of improved environmental quality to have additional
impacts through reducing demand for healthcare goods and services and therefore mitigating the
economic inefficiencies of subsidies in those markets.
In addition to the previously described interactions, the introduction of environmental quality into a
dynamic setting may lead to additional interactions with respect to the way in which changes in mortality
risk affect savings decisions, which may interact with other pre-existing capital taxes. The specific
magnitude and ultimate impact of the interactions on the net benefits of air regulations in the United
States is an empirical question that requires additional study. However, it is possible that benefit-side
general equilibrium interactions could be of first order importance for understanding the net welfare
impacts of environmental policies. Though studying those effects in an empirical setting may have
numerous challenges including the definition of appropriate closure rules in a dynamic setting..
4 Additional Factors Affecting the Incorporation of Benefits into CGE
Models
The research reviewed in the previous section provides an important step forward in estimating the
economy-wide benefits of environmental policies. However, there remain a number of limitations and
challenges associated with estimating a more comprehensive value for the economy-wide effects of
improvements in air quality. This section considers a number of these areas.
4.1 Valuing Mortality Risk Reductions
As discussed in Sections 3.1 and 3.2, to date most CGE modeling studies that have examined the
macroeconomic implications of human health have represented the proximate effects of changes in
32
mortality risks through changes in the representative individuals’ time endowments. The welfare effect
of a change in time endowment on the margin will be driven primarily by the wage rate. As noted by
Matus et al. (2008), though, the welfare impact will diverge from losses in expected income over time due
to the compounding of other general equilibrium effects captured by the models. For analyses examining
the impact of changes in premature mortality on economic variables such as gross domestic product
(GDP), representing mortality risk changes by modifying the time endowment of the representative agent
may be a reasonable starting point. However, in the context of applied welfare analysis, measuring the
benefits of changes in premature mortality as a reduction in the time endowment is closely related to the
“human capital approach” to valuing loss of life. In general, this approach involves estimating the benefits
of mortality risk reductions using the change in an individual’s expected remaining lifetime earnings
(Linnerooth, 1979; Rosen, 1988; Johansson, 1995; Cropper, 2000). The human capital approach was in
common use before Schelling (1968) and Mishan (1971) ushered in the application of willingness to pay
measures for mortality risk reductions in benefit-cost analyses. The use of WTP measures is now standard
practice, and as discussed in Section 2.2, is the main basis for EPA’s benefit-cost analyses of the human
health impacts of regulations.
In light of the similarities between the human capital approach and the time endowment approach that
has been used in most CGE models to date, in this sub-section we review the distinctions between the
human captial approach and contemporary willingness to pay approach to valuing mortality risk changes.
We also consider the quantitative difference between EPA’s (2011) benefit estimates (based on
willingness to pay) and CGE results (based on changing the representative agent’s time endowment) in
light of these distinctions.
To begin, Schelling’s (1968) early discussion of the conceptual differences between the human captial
approach and the willingness to pay approach is still apt and worth revisiting:
“People get hung up sometimes on the apparent anomaly that if a person would yield 2 percent
of his lifetime income to eliminate a 1-percent risk of death, he'd have to give up twice his entire
lifetime income to save his own life—which he cannot do if his creditors are on their toes. But he
doesn't have to. I'd pay my dentist an hour's income to avoid a minute's intense pain—even to
prevent somebody else's pain—without having to know what I'd do if confronted with a lifetime
of intense pain. This is why the 'worth of saving a life' is but a mathematical construct when
applied to an individual's decision on the reduction of small risks... Let me guess. If we ask people
what it is worth to them to reduce by a certain number of percentage points over some period
the likelihood that they will die, they will find it worth more than that percentage of the
discounted value of their expected lifetime income. Arithmetically, if we tell a man that the
likelihood of his accidental death over the next three years is 9 percent and we can reduce this to
6 percent by some measure we propose, and ask him what it is worth to reduce the probability
of his death by 3 percent over this period (with no change in his mortality table after that period),
my conjecture is it is worth to him a permanent reduction of perhaps 5 percent, possibly 10
percent, in his income...”
The quantitative difference between the value of mortality risk changes estimated under the human
capital and willingness to pay approaches can be very large. For example, the present value of future
33
earnings for a middle-aged worker paid the average wage in the U.S. is around $0.75 million (2008 U.S.
$).17 This figure can be compared to empirical estimates of the VSL based on hedonic wage and stated
preference studies. Central estimates of the VSL from recent meta-analyses of both revealed and stated
preference studies are between $2 and $11 million (2008 U.S. $) (Cropper et al., 2011), and EPA’s central
estimate of the average VSL is $7.9 million (2008 U.S. $) (US EPA, 2010).18 Therefore, contemporary central
estimates of the willingness to pay for marginal mortality risk reductions are up to an order of magnitude
larger than the average remaining lifetime earnings for a middle-aged U.S. worker.
This large quantitative difference raises two important issues for the representation of health benefits in
CGE models. The first is that representing expected changes in mortality risks as a change in the
representative agent’s time endowment may significantly underestimate the welfare effect of air
pollution regulations in a CGE analysis. As discussed in detail in Section 3.1, EPA (2011) used a CGE model
to examine the macroeconomic impacts of the CAAA. Two scenarios were examined: a “costs only”
scenario, in which the impacts of compliance expenditures alone were included, and a “labor force-
adjusted” scenario. The latter scenario included both compliance expenditures and beneficial health
impacts (specifically, PM induced mortality risk reductions and ozone and PM related morbidity for the
working population), where these health impacts were primarily modeled as a change in the
representative agent’s time endowment. EPA’s goal in that study was to examine the macroeconomic
implications of the CAAA, not to conduct an applied welfare analysis. Nevertheless, the benefit study and
the CGE analysis conducted by EPA (2011) provides an additional illustration of the quantitative difference
between benefit estimates inferred from adjusting the expected time endowment in a CGE model and a
willingness to pay measure for marginal mortality risk reductions based on the VSL. If one takes the
difference between the estimates of Hicksian equivalent variation for the “costs only” and “labor force-
adjusted” cases as an alternative measure of the benefits of the CAAA (Smith, 2012), this leads to an
estimate of $65 billion (2006 US $) in 2010 (US EPA, 2011 Tables 8-7 and 8-8). In the same study, EPA
reported a central estimate of the annual WTP for the same health endpoints included in the CGE analysis
17 This figure was computed based on an individual who is 40 years of age and earns $43,168 (2008 U.S. $) per year,
which was the annual mean wage in the U.S. in 2013 (http://www.bls.gov/oes/current/oes_nat.htm#00-0000). The
individual’s real wage is assumed to remain constant for the remainder of her working career (i.e., her nominal wage
grows at the rate of inflation), she will retire at age 65, and she discounts future consumption at a rate of 3% per
year. Finally, we assume that the individual is subject to a lifecycle mortality risk profile that matches average
mortality risks among the U.S. population. The individual’s age profile of mortality rates is estimated using a
Gompertz function, where her probability of death at age 𝜏, 𝑚𝜏, increases exponentially with age, i.e., 𝑚𝜏+1 =
(1 + 𝜂)𝑚𝜏. A good fit to average mortality rates among the general U.S. population ages one and above (Xu et al.
2010) can be achieved using 𝑚0 = 0.0001175 fatalities/yr and 𝜂 = 0.0780. The individual’s expected remaining
lifetime earnings, or “human capital” (HC), are then calculated as 𝐻𝐶 = ∑ 𝑠𝑎,𝑡(1 + 𝑟)−(𝑡−𝑎)𝑦𝑡𝑅𝑡=𝑎 , where 𝑠𝑎,𝑡 =
∏ (1 − 𝑚𝜏)𝑠𝑎,𝑡𝑡𝜏=𝑎 and 𝑦𝑡 is her real wage at age t.
18 Worth noting is that under the same assumptions about the average middle aged U.S. worker used in Footnote X,
the conjecture of Schelling (1968) implies a VSL between $2.6 and $5.2 million, which is comfortably within the range
of estimates summarized by Cropper et al. (2011).
34
equal to $1.2 trillion (2006 US $) in 2010 (US EPA, 2011 Table 7-1). The difference between EPA’s central
benefit estimates and the (superficially) analogous welfare measure from the CGE analysis is consistent
with the difference between people’s marginal WTP for mortality risk reductions and the change in their
expected remaining lifetime earnings (as summarized above), accounting also for the fact that the time
endowment was adjusted only for working individuals (i.e., health improvements for children, retired
individuals, or others not included in the labor force are absent from the CGE analysis).
The second issue raised by the large quantitative difference between the human capital and WTP
approaches is what the difference means for modeling behavior. In models that include some degree of
foresight and that allow for intertemporal wealth transfers, changes in the expected future time
endowment or mortality risks will influence consumption and labor supply patterns. If the agent’s WTP
for a mortality risk reduction is in fact an order of magnitude larger than the expected value of the change
in the agent’s time endowment, we might also suspect that the agent’s behavior would be notably
different if health impacts are modeled explicitly as a change in mortality risk rather than implicitly as a
change in the time endowment. If so, then this issue may have important implications not only for the use
of CGE models to estimate the welfare effects of environmental regulations but also their ability to
accurately represent the effects of environmental regulations on GDP and other macroeconomic
outcomes.
4.2 Health Risks to Children
In addition to the mortality risk reductions discussed in the previous section, air quality improvements
also impact the health of children, and those impacts are not necessarily the same as those experienced
by adults for at least two reasons. First, children are in some cases more highly exposed to environmental
contaminants relative to adults due to normal childhood activities and behavior patterns. According to
America’s Children and the Environment (US EPA, 2013a), “children generally eat more food, drink more
water, and breathe more air relative to their size than adults do, and consequently may be exposed to
relatively higher amounts of environmental contaminants.” Second, children may be more vulnerable to
environmental contaminant exposure compared to adults since they are still developing physically and
may therefore, be more easily harmed.
Executive Order 13045, Protection of Children from Environmental Health Risks and Safety Risks, directs
all Federal Agencies, including EPA, to make assessing children’s environmental health and safety risks a
high priority. When suitable exposure-response relationships can be identified, EPA has quantified child-
specific benefits in regulatory analyses. As noted in Table 1 the child-specific health impacts that have
been incorporated into standard benefit-cost analyses associated with changes in exposure to air
pollutants include PM-related infant mortality and acute bronchitis, ozone related hospital admissions for
respiratory causes and school absences, and SO2-related asthma exacerbation and acute respiratory
illness.
Fully capturing the benefits of reducing children’s health risks can be challenging. As noted previously,
standard practice in monetizing adult environmental health risk reductions relies on individual WTP
estimates for ex ante risk reductions or cost of illness estimates as a proxy when WTP estimates are
unavailable. However, there are a number of reasons why these measures may introduce bias when
35
applied to children, including differences in risk preferences (e.g., risk aversion, involuntariness of
exposure), duration of health outcome, and remaining life expectancy (US EPA, 2003). Furthermore, the
standard practice of relying on individual, own-preferences for risk reducing behaviors may not be
appropriate when considering children since it may not be possible to assume that they are well informed
and rational economic agents. As such, the literature on assessing the benefits of reducing children’s
health risks has evolved around household decision-making models, with parents at the helm.
Gerking and Dickie (2013) reviewed the existing empirical literature (including both stated preference and
safety product valuation studies) and categorized studies based on the conceptual model of household
production considered: (1) single parent households in which only the child faces a health risk; (2) single
parent households in which both the parent and the child face a health risk; and (3) two parent households
in which both the parents and the child face health risks. They found that parents’ marginal WTP for a
one-unit reduction in health risk to their child generally exceeds their WTP for the same absolute risk
reduction to themselves. A number of competing explanations exist for this finding beyond parental
altruism, for example, the positive perception that protective actions might garner from others. It has also
been suggested that this result could be due to differential parental discounting of latent risks over latency
periods that differ in length depending on life-stage.
Gerking and Dickie (2013) noted two important limitations with existing studies that value changes in
children’s health. First, even though mortality and morbidity risks are intertwined (US EPA, 2010b),
valuation of morbidity risks and mortality risks are generally modelled separately as if independent,
thereby leading to a potential double-counting of benefits.19 Second, most empirical models of valuing
changes in children’s health are based on a unitary framework that assumes a single set of preferences
across parents in the household, whereas family dynamics may lead to more complicated decision-making
scenarios.
Adamowicz et al. (2014) explored alternative models of household decision-making, which allowed them
to address this second challenge by modeling households as a collective in which parents may have
different preferences and potentially different levels of decision-making power. In this setting, parents
may be assumed to work cooperatively or not. Using stated preference data collected from 432 matched
pairs of parents’ they empirically test these various models. They found Pareto efficiency in health
resource allocations across the household and an inconsequential effect of redistributing the household’s
budget across members of the household on the marginal willingness to pay for health risk reductions for
the child. This result lends support to both the unitary and cooperative models of household decision-
making.
Gerking et al. (2014) developed an integrated model of mortality and morbidity. The model extended the
standard one-period expected utility model with two states (alive and dead) to include a third state, the
probability of being ill, with the utility experienced at this third state falling between the other two. The
model adopted a family perspective in which the parent makes decisions about risk exposures for the
child and herself and spanned multiple periods allowing for differences in latency across parents and
19 This shortcoming is not exclusive to children’s health valuation but more broadly to health valuation at all ages.
36
children for the health outcomes. Using stated preference data collected from two separate surveys, one
on leukemia risks and the other on skin cancer risks, the authors confirmed two important findings from
their conceptual analysis of the model. First, parents consider morbidity risk and conditional mortality risk
as perfect substitutes. That is “in both studies, total health benefits of reduced health risks for children
are entirely captured by the marginal willingness to pay to reduce unconditional mortality risk.” Second,
a parent’s willingness to pay for risk reductions to herself is not the same as that for the same risk
reduction in her child, but neither is it systematically lower or higher. Further, they find a potential for
double counting if separate morbidity benefits are added to the total benefits calculation and that “adult
VSL values may be a questionable guide to corresponding VSL values in children.” These findings, taken
together, point to the need for additional studies of WTP for children’s health risk reductions.
While the approaches discussed above hold promise, WTP estimates for children’s environmental health
risks are often limited regardless of which model is used. While EPA has identified adequate WTP
estimates for some health outcomes, and applies these routinely to child-specific endpoints in its benefits
calculations for reduced PM exposures (see Table 3), these WTP estimates are rarely generated specifically
for children, and so, do not capture the differences in WTP values described above and the nuances
surrounding household decision making. As reported in section 2.2, cost of illness estimates are often
used as a proxy for the missing WTP values in benefit-cost analyses. To the extent that age-specific cost
of illness estimates can be derived, these are applied where appropriate. Otherwise, adult values – be
they based on WTP or COI – are transferred to children’s health endpoints in the monetization step in
benefit cost analyses when child specific values are lacking.
As discussed in Section 3, previous studies on the benefits of air quality improvements have incorporated
changes in children’s morbidity and mortality risks to varying degrees using different methodologies. The
EPA (2011) analysis, while providing a robust qualitative and in some cases quantitative partial equilibrium
analysis of the benefits of reducing children’s health risks through air quality improvements, only provided
a limited inclusion of these impacts in the CGE analysis. Specifically, the only benefits associated with
children’s health impacts included in the CGE analysis were reduced school loss days representative of the
lost time endowment for the working parent required to care for the child. No benefits representative of
the willingness to pay by, or on behalf of, the children themselves for their health risk reductions are
captured in the model.
As noted in Section 3.2, the studies using the EPPA-HE model (Matus et al., 2008; Nam et al., 2010; and
Matus et al., 2012), accounted for the value of lost non-work (leisure) time associated with morbidity
rates in children due to air pollution exposure. Based on the U.S. wage profile by age, lost leisure time for
cohorts below the working age due to morbidity impacts are valued at one-third the wage rate. Further,
the model incorporated values for market and non-market effects of illness (e.g., pain and suffering and
other associated losses) by apportioning WTP estimates among three components: medical expenditures
(demand for medical services), lost work time (reduction in labor force), and “damages beyond these
market effects” (which as a proxy are considered a loss of leisure). Whereas the valuation estimates used
by Matus et al. (2012) do not appear to include any that were derived specifically for children (see Holland
et al., 1998), those used by Nam et al. (2010) and Matus et al. (2012) were expanded to include a few
37
child-specific WTP estimates, including one for respiratory symptoms in asthmatic children (see Bickel and
Friedrich, 2005).
With this approach, the effect of children’s environmental health risks will be below that for adults for the
same outcomes. This is in contrast to the recommendations by OMB in Circular A-4 (OMB, 2003) on the
valuation of children’s health risk changes in benefit cost analyses, where OMB recommends that agencies
use values for children in benefit-cost analyses that are “at least as large” as those for adults for the same
outcome, barring any compelling evidence to the contrary.
The potential importance of accounting for children’s health risks in CGE models follows from the fact that
in a dynamic setting, changes in the health risks faced by children can have longer-term implications for
the economy than comparable health risks faced by adults. Nevertheless, as noted above there are many
challenges associated with developing partial equilibrium WTP measures for improvements in the health
risks facing children. Given the challenges of incorporating the changes in adult mortality risks in CGE
modeling, as described in the preceding section, and the challenges in developing partial equilibrium
estimates for childrens’ health benefits, there are many open questions about the appropriate methods
for children’s health risk changes in CGE models.
4.3 State Dependent Utility Functions
A common choice within CGE models is to represent households as maximizing homothetic utility
functions, such as the constant elasticity of substitution or Cobb-Douglas functions. These functions are
convenient because they are easily aggregated and can be represented with a relatively parsimonious set
of parameters that can be calibrated using existing market data. However, these functions impose
restrictions on consumer behaviors that may be unrealistic; specifically, the elasticity of substitution
between final goods is fixed and budget shares for final goods are determined by relative prices and not
affected by income. Furthermore, in these typical specifications, health does not play a direct role in
affecting households’ well-being or preferences, as represented by the marginal utility of consumption or
leisure and the elasticity of substitution between labor and leisure or between other final goods.
As discussed in Section 3, past studies that have incorporated the health benefits of air pollution
regulations in CGE models typically assume that health affects the utility of representative households
only through its impact on medical expenditures and the time endowment. While this approach is
attractive due to its tractability, it does not take into account the possibility that health status may either,
directly affect utility (independent of effects on consumption), or may be a complement to consumption
and leisure.
Various hypotheses about the role of health status in the utility function have been examined analytically
and to a limited extent tested with empirical data in the health economics literature (Lillard and Weiss,
1997; Finkelstein et al., 2009; Viscusi and Evans, 1990; Sloan et al., 1998). This relationship is commonly
discussed in terms of “health state dependence,” meaning “the effect of health on the marginal utility of
a constant amount of nonmedical consumption” (Finkelstein et al, 2009). While being in an undesirable
lower health state must lower the level of utility for an individual at any given level of consumption, the
sign and magnitude of the effect on marginal utility (the derivative of utility with respect to consumption)
38
is not known a priori. If consumption goods are predominately complements to health status, then
negative state dependence might be likely, whereby a deterioration in health is accompanied by a
reduction in the marginal utility of consumption. This could be the case where a deterioration in the health
state inhibits the enjoyment of consumption such as the flu reducing the enjoyment of a vacation.
However, in the case where final goods and services can substitute for good health there may be positive
dependence where the marginal utility of consumption increases with deteriorating health.20 For
example, reduced health states that inhibit non-market production such as cooking increase the marginal
utility of prepared meals, or a broken hip may require increased utilization of taxis. The difference
between these two cases is depicted in Figure 3.
To examine health state dependence, Finkelstein et al. (2013) used panel data on health status and
subjective proxies of utility for the elderly and near-elderly to model the way in which utility proxies
change across health shocks and consumption levels. They found negative state dependence where the
marginal utility of consumption is 10-25% percent lower for individuals in a health state characterized by
a chronic disease (e.g., hypertensions, diabetes, cancer) relative to a healthy state, holding the level of
consumption constant.
Sloan et al. (1998) used results from a stated preference survey to estimate the scaling coefficients for log
utility functions that differ between a healthy state and one with multiple sclerosis. They estimated that
the marginal utility of consumption in the health state with multiple sclerosis was 92% lower than in the
healthy state for the general population. However, when the sample is restricted to those with multiple
sclerosis the marginal utility in the unhealthy state was only 33% lower.
Viscusi and Evans (1990) also used stated preference survey data on compensating differentials for job
related injuries in the chemicals sector to test the health state dependence of the utility function. They
found evidence of negative state dependence with marginal utility in the injured state being 7% to 23%
lower than in the pre-injury state. These results were driven by the potential for more severe injuries
lasting more than a month. When the sample was restricted to more minor and temporary injuries they
found some evidence of positive state dependence.
Murphy and Topel (2006) calculated the value of improvements in health and life expectancy in the U.S.
from 1970 to 2000. They argued that improvements in life expectancy are important for such valuation
because of the additional welfare gained from the additional expected years of life, and the improvements
in quality of life from health improvements which “raise utility from given amounts of goods and leisure.”
This prior was implemented in their framework through the imposition of negative state dependence by
multiplying the utility by the current health state measured from zero to one. At its peak effect, Murphy
and Topel (2006) suggested that general health improvements over the 30 year time period increased the
marginal utility for 55 year old males by nearly 10%.
20 Specifically, the sign of state dependence requires the cross partial of utility over health and non-medical
consumption to be non-zero, such that for a negative cross partial (positive state dependence) the two are
Edgeworth-Pareto complements (Samuelson, 1974).
39
a: Marginal Utility Declines with Sickness
b: Marginal Utility Increases with Sickness
Figure 3: Health State-dependent Utility Functions [Source: Finkelstein et al., (2013)]
Similar to Viscusi and Evans (1990), Evans and Viscusi (1991) used stated preference survey data on
compensating differentials for consumer product risk to examine the health state dependence associated
with minor and temporary health effects. They assumed a logarithmic utility function and developed an
estimation procedure to determine whether the responses exhibit negative state dependence or
equivalence between the negative health effect and a loss in consumption. The restricted model does not
allow for the possibility of positive state dependence. Evans and Viscusi (1991) rejected negative state
dependence for this class of minor health effects.
Lillard and Weiss (1997) examined health state dependence by studying the effect of health state shocks
on dynamic consumption patterns of retired households between 1969 and 1979. The authors estimated
the parameters of a stochastic dynamic programming framework for household welfare maximization
using consumption and health data from the Retirement History Survey. Both the parameter estimates
40
and subsequent simulations using their model showed strong positive state dependence for married
couples where one member enters a lower health state, but failed to find evidence of health state
dependence for single female households.
Edwards (2008) studied the potential rationale for a continued decline of household exposure to financial
risk after retirement. In particular, Edwards (2008) assessed whether this investment pattern is better
explained by nearing medical expenditures or due to positive state dependence that would increase
marginal utility in the increasingly more likely event of being in a lower health state. Using data from 1992
to 2000 from the Health and Retirement Survey and the Study of Assets and Health Dynamics Among the
Oldest Old, the author estimated the parameters of a stylized model of portfolio choice that disentangles
the effect of health state dependence and risk aversion, including the effect of potential out of pocket
medical expenses. Edwards (2008) found evidence of positive state dependence, though noted some
caution is warranted when interpreting the magnitude of the results due to the stylized nature of the
model.
In a review of studies examining health state dependence Finkelstein et al. (2009) noted that “a priori, the
sign (let alone the magnitude) of any health state dependence is ambiguous” and that “we have relatively
little empirical evidence on health state dependence, and what we do have is inconclusive in both sign
and magnitude.” However, stylized numerical exercises have found that even moderate amounts of
health state dependence can have large effects on the optimal level of lifecycle savings (Finkelstein et al.,
2009). Moderate amounts of health state dependence may also have a notable impact on the welfare
change associated with changes in air quality. Therefore, while there may remain serious concerns about
the ability to calibrate any potential approach of incorporating health state dependence into studies on
the benefits of air regulations using CGE models, the effect of omitting such impacts has the potential to
be significant.
The previous literature on health state dependence offers a few potential approaches, which may be
useful for incorporate health state dependence into a CGE framework. Health state dependence could be
represented by having multiple representative agents that differ by health status, similar to the way in
which some CGE models represent different household demographics. As in Evans and Viscusi (1991) and
Murphy and Topel (2006), it could be assumed that total utility is simply scaled between the types of
representative households depending on their health state, with lower health states being affected by a
multiplier less than one. Such an approach may provide a tractable solution as the literature provides
empirical estimates of this scaling factor in some circumstances. However, this approach would
necessarily impose negative state dependence, such that the marginal utility associated with consumption
of all final goods decreases with health status. An alternative approach would be to adjust the productivity
of particular final goods and leisure between the representative households depicting various health
states. This approach would provide the ability to allow final goods and leisure to vary with respect to
whether they are complements or substitutes for health status. This approach may provide a more
realistic depiction on the effect of health status and would leverage the abilities of CGE models. However,
calibration of such an approach may be beyond the scope of the current empirical literature on health
state dependence.
41
While such approaches may be tractable within typical CGE frameworks, they would likely assume that
transitions between health states are complete surprises to households. In reality, consumers are aware
ex ante of the potential risk and potential to transition between health states, and will behave accordingly.
An environmental regulation that lowers the risk of entering a lower health state may cause households
to adjust their behavior. As shown by Finkelstein et al. (2009) and Edwards (2008) such changes may
include adjustments to savings behavior that could have important impacts within a CGE setting if the
effect is non-marginal.
4.4 Indirect Impacts on Human Health
The primary pathway for air regulations to impact human health is through improvements in air quality
as discussed in Section 2. However, there may also be indirect impacts on human health through other
pathways that may also be of interest. One channel by which human health may be indirectly affected by
air regulations is through a decrease in the real wage, due to an increase in prices of goods and services
relative to the wage rate. An income effect may cause households to reduce their consumption of
healthcare goods and services, which in turn may reduce the health state for the household. There may
also be a substitution effect whose directional impact on households’ purchases of healthcare goods and
services, and in turn their health state, will depend upon on the relative effect of the regulation on
healthcare prices compared to other final goods and services. Furthermore, improved health through the
direct impacts of the regulation may reduce the demand for healthcare goods and services, thereby
lowering the price and potentially cause households to substitute towards such goods and services
improving their provision of health.
Ex ante, the ultimate impact of these competing indirect effects on household health is ambiguous.
However, under the assumption of perfect and complete markets, the value of this change in health to
society will be captured in a partial equilibrium welfare measure, as prices will capture the opportunity
cost to society of compliance, inclusive of any household indirect health effects. In the more realistic case
where distortions exist in the economy, CGE models are particularly adept at capturing changes in relative
prices and therefore, it may be worth considering whether these models can offer improvements in
capturing the indirect effect of air regulations on households’ health. In their review article of tools for
comparative analysis, Hofstetter et al. (2002) suggest that tracking such indirect effect may very well
require data-intensive general equilibrium models. Implicitly, this class of models already captures the
demand by households’ for healthcare, which is conditional on the welfare benefits this type of
consumption provides through its effect on health. As such, social cost estimates generated by CGE
models already implicitly take into account the potential indirect impacts the costs of complying with an
air regulation would indirectly have on households’ health.
However, improvements may be possible to more robustly capture households’ demand for healthcare
goods and services. This may require directly modeling the way in which consumption of those goods and
services impact health and ultimately household welfare. As noted in Section 3.2, there are a limited
number of studies that have considered the endogenous provision of health through the consumption of
healthcare. However, additional research may be warranted given the importance of household
heterogeneity (across multiple dimensions) and pre-existing distortions in the healthcare market, both of
42
which may be important for understanding the ultimate effect on the endogenous provision of health by
households.
Air regulations may also have indirect impacts on human health through their interaction with labor
markets. For example, the reallocation of labor across the economy in response to an environmental
policy may result in transition costs borne by individual households in moving to the new policy
equilibrium. These transition costs could potentially affect their budget constraint and in turn, their
endogenous provision of health, though the modeling of such impacts would appear to be subject to the
challenges noted above and those associated with capturing transition costs in a CGE model, as discussed
in the whitepaper on Economy-Wide Modeling: Social Cost and Welfare. This reallocation in the labor
market may also shift the distribution of on-the-job risks across the workforce, which could be relevant
for understand the net impact of labor market changes on household health.
4.5 Air Quality and Labor Productivity
Outdoor workers may be at increased risk of experiencing health impacts associated with air pollution
due to their increased exposure. As previously discussed, such exposure may lead to effects on the
extensive margin as the quantity of labor supplied is reduced due to workdays lost to illness or premature
mortality. In addition to these effects, in some cases the elevated exposure for outdoor workers may also
lead to less visible effects on the intensive margin as poor air quality could impair worker productivity and
therefore reduce the amount of “effective labor” supplied.
There is evidence that in some cases outdoor workers may experience temporary health impacts
associated with short-term increases in exposure to air pollution. For example, short-term exposure to
elevated ozone concentrations has been shown to notably affect lung function in healthy adults (US EPA,
2013b). Specifically, decreased lung function due to ozone exposure has been observed in outdoor
agricultural (Brauer et al., 1996) and forestry (Hoppe et al., 1995) workers. Just as such, health impacts
have been found to reduce the speed of hikers (Korrick et al., 1998), they may have an effect on the
productivity of outdoor workers.
In an early examination of the subject, Crocker and Horst (1981) used a small case study in southern
California to examine the impact of ground-level ozone concentrations on agricultural workers.
Specifically, over the course of a single season they examined the productivity of 17 citrus pickers who
were paid on a piece-rate basis, i.e., a fixed rate was paid per box of fruit picked. Air quality was found to
be a significant determinant of productivity for a subset of the workers studied. In a more recent study,
Zivin and Neidell (2012) more thoroughly examined the impact of ground-level ozone concentrations on
the productivity of agricultural workers in California’s Central Valley. Using data on the output of workers
paid through piece-rate contracts they observe that a 10 parts per billion decrease in ambient ozone
concentration for the area was associated with a 5.5% increase in worker productivity.21 While this result
is based on agricultural activity in a particular location, they note that 11.8 percent of the U.S. labor force
is associated with industries with regular exposure to outdoor conditions.
21 The average concentration in the sample was 50 ppb and the current federal NAAQS is 75 ppb.
43
Given the appropriate sectoral and possibly regional disaggregation, the mechanics of introducing labor
productivity shocks conditional on air quality into a CGE model could be relatively straightforward. The
larger challenge may be the availability of estimates that would allow for the mapping of air quality levels
to sector-specific labor productivity. Such estimates are essential for appropriately calibrating the labor
productivity shock that would be introduced into the model conditional on air quality changes. While
there has been substantial work examining the influence of air quality on physical health outcomes (e.g.,
lung function), very little is known about the influence of such health outcomes on labor productivity.
4.6 Material Damages from Atmospheric and Deposition Effects
As briefly noted in Section 2.1, atmospheric and deposition effects of air pollution can lead to materials
damages. These damages can be broken into three categories or primary interest: corrosive effects on
commercial and industrial buildings, household soiling, and corrosive and soiling effects on structures of
cultural and historical significance. Criteria air pollutants can have corrosive effects on some materials
leading to increased rates of depreciation. Materials such as paint, galvanized steel, mortar, and stone
used to construct walls, roofs, fencing, and window trim (etc.) may all be susceptible to damages from
acid deposition effects, resulting in reduced time intervals between repairs or replacement. Estimating
the impacts of acid deposition changes because of air regulations is a complicated problem that requires
estimating the materials response to acid deposition, modeling the deposition effects at an appropriately
fine spatial scale, and estimating the spatial distribution of sensitive materials. Similarly, increases in
atmospheric CO2 concentrations lead to increased depreciation rates for some materials such as concrete
(Saha and Eckelman, 2014). If all of this information is available, which is not necessarily the case for most
regulations, it may be possible to aggregate the estimates based on the proportion of the capital stock
affected to regional impacts on the capital depreciation rate in a CGE analysis.
Household soiling refers to the accumulation of dirt, dust, and ash on exposed residential structures.
While some recent studies have sought to estimate this benefits category (Yoo et al., 2008), the majority
of the estimates in the literature have been deemed too narrow in their spatial scope or too outdated for
use in current analyses of national air quality improvements (US EPA, 1998).
Improvements in air quality will also reduce soiling of, and damages to, buildings and structures of cultural
and historical significance. Society’s willingness to pay for preventing such damages through
improvements air quality may be significant (Grosclaude and Soguel, 1994), in part due to the non-
substitutability intrinsic to such assets (Kling et al., 2004). However, the studies estimating the benefits of
preserving historical and cultural structures are naturally very site specific (e.g, Willis, 1994; Chambers et
al., 1998) limiting the potential for benefits transfer approaches. Furthermore, it is unclear how such
values would be incorporated into a CGE framework, though the benefits may have a strong impact on
the relative demand for regional tourism services.
4.7 Air Quality and Productivity in the Agricultural and Forestry Sectors
Air pollution has the potential to affect vegetation in multiple ways, including induced foliar injury,
reduced annual growth rates, and induced root loss (US EPA, 2007). Of particular concern is ground-level
ozone that results from emissions of nitrogen oxides and volatile organic compounds (VOCs). These
44
impacts on vegetation can have meaningful direct effects on the agriculture and forestry sectors of the
economy, in addition to non-market welfare effects that may also affect economic activity as discussed in
Section 3.3.
Visible foliar injury can reduce the desirability of leafy crops, (e.g., spinach, lettuce) and ornamentals (e.g.,
petunias, geraniums, poinsettias). The discoloration effects and early shedding of leaves, as a result of air
pollution, can also have notable welfare and economic effects in areas where foliage is highly valued (e.g.,
scenic vistas, fall foliage). Elevated levels of ozone can reduce the overall health of plants through
inhibiting their physiological processes (US EPA, 2007). These impacts can directly affect the productivity
of agriculture and forestry sectors. Conceptually, the inclusion of the impacts of air pollution on the
productivity of land used in the agriculture and forestry sectors may be straightforward, but in practice,
there are a number of challenges. Air pollution, such as ground-level ozone, is not uniform across space,
and its effects are not uniform across plant species. Therefore, there are potential adaptation
opportunities for the economy through substitution. In highly aggregated CGE models, it may be difficult
to immediately capture these substitution possibilities, both spatially and sectorally. Previous analyses
conducted by EPA, such as the analysis of the benefits and costs of the Clean Air Act from 1990 - 2020 (US
EPA, 2011), have considered the impacts of emissions reductions on agricultural productivity, though in
detailed partial equilibrium settings that more completely represent some of the substitution possibilities.
Incorporating results of such modeling exercises, or the details that allow the estimation of net
productivity shocks, is similar to the linkages topic considered in the whitepaper on Economy-Wide
Modeling: Social Cost and Welfare, and answers to charge questions on that topic may be applicable here.
For the impact of improved air quality on plant health and its implications for non-market services, the
challenges are similar to those previously described in Section 3.3.Spatial Challenges in Benefits
Assessment
The heterogeneous impact of air quality improvements across space presents unique challenges for
assessing the benefits of environmental policy in general, and potentially the use and/or interpretation of
CGE-based benefits assessments in particular. In this section we focus on two areas, the degree of spatial
heterogeneity in the benefits of air quality improvements relative to the spatial resolution of typical CGE
models; second, and the potential first order welfare effects associated with re-sorting in housing markets
in response to exogenous changes in the provision of public goods (i.e., air quality).
4.8 Spatial Resolution of Benefits
The impacts of regulations designed to improve air quality have effects that can differ significantly across
space due to a number of crucial factors. First, affected sources are not distributed uniformly across space,
and in many cases, the change in emissions is not uniform across sources. Second, the relevant changes
in air quality may not be associated with the pollutants that are directly emitted but are instead formed
through secondary processes. For example, tropospheric ozone forms because of a complex series of non-
linear chemical reactions of precursor pollutants, such as volatile organic compounds and nitrogen oxides.
Therefore, the impacts associated with changes in emissions depend on the dispersion and transport of
the pollutants prior to such reactions, and are influenced by spatial heterogeneity in weather patterns.
Finally, the ultimate impact on human populations depend on the correlation between the spatial
45
Figure 4: Predicted change in 2020 PM2.5 concentrations in EPA (2011)
distribution of air quality impacts and the spatial distribution of populations, demographics, and land uses.
As a result, the impacts of air regulations often need to be estimated a relatively fine spatial scale. Often,
estimates of the benefits are calculated at the spatial resolution of a 12 to 36 km grid. For example, Figure
4 presents the predicted changes in PM2.5 concentrations in 2020 modeled for EPA (2011) based on a 36
km grid.
It is readily apparent from Figure 4 that there is considerable variation in air quality impacts across the
United States. As previously noted, this information must be combined with information on the spatial
variation in population density and baseline incidence rates across the country, which leads to results for
any particular health endpoint that are very location specific. Introducing these results into a CGE model
will necessarily require aggregation up to the spatial scale at which economic activity is represented in the
model. This can be at the state (e.g., USAGE-R51, Dixon et al., 2012), multi-state regional (e.g., EMPAX-
CGE, US EPA, 2008), or national level (e.g., IGEM, Goettle et al., 2009). For comparison purposes, Figure 4
includes a delineation of the five multi-state regions in the EMPAX-CGE model. This figure therefore
demonstrates the level of aggregation that was used in EPA (2011) to translate the estimated health
46
impacts into their corresponding effects on the regional time endowments and household healthcare
expenditures for inclusion in the CGE model. Given the assumption of perfectly mobile capital and labor
within the CGE model regions, this aggregation could lead to cases where, because of the regulation,
households in one part of a region supply additional labor to industries located in a different part of the
region. Therefore, the degree to which this aggregation affects the results will depend on the degree to
which there is any potential mismatch between the air quality patterns and economic activity within
regions. This issue of aggregation may also affect the results of the CGE model in other ways, for example
if the deviation of local healthcare costs from the regional average are correlated with spatial distribution
of health impacts. The extent to which these issues might affect the results is unknown.
4.9 Spatial Sorting and Welfare Effects
The approaches to modeling the benefits of air quality improvements discussed in Sections 2 and 3 are
based on the assumption that either the spatial distribution of the affected population remains constant
or shifts based on an exogenous projection are independent of changes in air quality. However, the change
in air quality at a household’s pre-policy location may not be the relevant metric for assessing the impacts
of the policy on the household. Within housing markets households sort across neighborhoods based on
preferences for public goods, social characteristics, and proximity to labor and other markets. Regulations
that directly affect the supply of public goods, in this case clean air, can lead heterogeneous households
to re-sort as result of the changes in the available choice set.
Beyond the direct exogenous effect of the regulations, some amenities are endogenously determined,
creating a feedback between households and their environment. The initial shock associated with the
regulation induces sorting amongst households and leads to a potential redistribution of local amenities,
until equilibrium is reached within the market. Applications using equilibrium sorting models designed to
characterize preference heterogeneity across households have demonstrated that these may be first-
order effects for applied welfare analysis (Kuminoff et al., 2010).
A few studies have applied equilibrium sorting models to analyze the welfare impacts of environmental
policies. As opposed to the damage function approaches of Section 2 and the CGE approaches of Section
3, these methods focus on assessing the benefits of air quality improvements based on identifying the
effects of the exogenous policies that are capitalized in property values. Sieg et al. (2004) considered air
quality improvements in Southern California between 10 and 1995, when ozone concentrations were
reduced by 19% on average for the study area, with a range of 3% to 33% across individual communities.
Sieg et al. (2004) considered both static WTP estimates absent any household resorting and WTP
estimates with equilibrium sorting. At the school district level, they found WTP estimates with (costless)
re-sorting is on average 50% higher than the case where re-sorting effects are not considered.
This work was extended by Smith et al. (2004) to assess the ex-ante welfare impacts of the CAAA using
the forecast changes in air quality for 2000 and 2010 modeled in EPA (1999). They used the equilibrium
model of Southern California developed by Sieg et al. (2004) and simulated potential household resorting
and welfare impacts based on the changes in air quality for the region forecast by EPA modeling. When
allowing for resorting of households they found WTP estimates are on average around 10% higher
47
however, they found significant variation across neighborhoods: with equilibrium welfare impact
estimates up to six times greater than simulated estimates without re-sorting in some school districts.
Because sorting in the housing market will be based upon total environmental conditions including all
pollutants, equilibrium sorting models may have difficulty assessing the impacts of changes in the
atmospheric concentration of single but correlated air pollutants. Therefore, in can be difficult to use the
results of these models to assess the benefits of policies that change emissions of multiple pollutants, as
can been done with the damage function approach. This class of models to date have also primarily been
focused on local or regional analyses capturing single or tightly linked housing markets, as opposed to a
national level analysis. Furthermore, equilibrium sorting models may also have difficulty as currently
formulated in handling general equilibrium effects such as changes in labor supply and savings, as would
be captured in a general equilibrium setting. However, this class of models provides an ability to
incorporate other important aspects of the problem. First, they are able to capture a degree of
heterogeneity in household preferences and income effects, which may not be currently possible in CGE
models. Second, they allow spatial heterogeneity in certain household adjustments to policy changes at a
substantially finer level than CGE models, which represent the United States at a spatial resolution of
states in the finest case and a single region in the coarsest case. While directly modeling such aspects in a
CGE framework would be prohibitive, allowing for behavioral responses in terms of locational choice may
be important for understanding the impacts of changes in air quality on health points, such as changes in
mortality and morbidity risk, which are entered into the modeling. Additional linkages between modeling
frameworks may also offer opportunity for other improvements.
5 Concluding Remarks
EPA has extensive experience with modeling changes in air quality in response to regulations that act to
reduce the emission of pollutants into the atmosphere and subsequently estimating households’
willingness to pay for those air quality improvements using a damage-function approach. However, while
many of the estimates used for households’ marginal WTP for improvements in various endpoints are
derived, in part, by leveraging the influence environmental quality has on economic activity, these
feedback effects are not typically considered in the analysis of air regulations. As shown by the literature
discussed in Section 3, these economy-wide benefits of improvements in air quality can be significant. EPA
has some experience estimating the economy-wide impact that some of the health benefits borne by
improvements in air quality may have (US EPA, 2011), though, as detailed in the charge, technical
challenges and important methodological questions remain:
What are the main conceptual and technical challenges to representing the benefits of an air
regulation in an applied CGE framework, and what would be required to overcome them?
How do we reconcile more traditionally used estimates of WTP by individuals with estimates from
CGE models focused on overall welfare, and what insights does a CGE model provide when
benefits of air regulations cannot be completely modeled?
48
The EPA and outside researchers have worked to incorporate some of the human health benefits of
improved air quality in CGE models, by characterizing the main impact of changes in the incidence of
morbidity and pre-mature mortality as a change in the time endowment. However, the value to a
household of a change in their time endowment is not equivalent to their willingness to pay to reduce
morbidity or mortality risk. Furthermore, households’ preferences, as represented by their utility function
within CGE models, may be conditional upon the health state. These observations lead to questions of:
Is it technically feasible and appropriate to model mortality and morbidity impacts as a change in
the time endowment, and if not, what other approaches are available to incorporate mortality
and morbidity impacts into a CGE model?
Because of changes in their health and/or life expectancy, households may experience changes in
their relative preferences for goods and services. Could such changes be captured in a CGE model,
and under what circumstances would such modeling add value to the overall analysis?
While the predominant source of benefits from improvements in air quality are typically estimated to
derive from households’ willingness to pay to reduce the risk of premature mortality, there are a notable
number of other positive impacts associated reductions in air pollutions as detailed in Section 2. However,
CGE models may have value to add in the estimation of certain benefit categories and/or certain benefit
categories may be important for understanding the net economy-wide impacts of air regulations, leading
to the charge questions:
What approaches are available to incorporate additional effects of improved air quality in CGE
models, and what are the conceptual and technical challenges to incorporating them?
What techniques are available for incorporating impacts of improved air quality on non-market
resources in existing CGE models, and what are the particular challenges to incorporating non-
market benefits into a general equilibrium framework (e.g. non-separability)?
49
References
Abt Associates (2008). Environmental Benefits Mapping and Analysis Program (BenMAP) User’s Manual.
Report Prepared for Office of Air Quality Planning and Standards, U.S. Environmental Protection Agency,
Research Triangle Park, NC, Neal Fann, Project Manager.
Adamowicz, W., Dickie, M., Gerking, S., Veronesi, M., and Zinner, D. 2014. “Household Decision Making
and Valuation of Environmental Health Risks to Parents and Their Children.” Journal of the Association of
Environmental and Resource Economics 1(4): 481-519.
Agency for Healthcare Research and Quality. (2000). HCUPnet, Healthcare Cost and Utilization Project.
Rockville, MD.
Antoine, B. Gurgel, A., and Reilly, J. 2008. “Will Recreation Demand for Land Limit Biofuels Production?”
Journal of Agricultural & Food Industrial Organization 6(2):1-29.
Bickel, P., and Friedrich, R. 2005. ExternE: Externalities of Energy, Methodology 2005 Update. European
Commission, Community Research.
Berger, M., Blomquist, G., Kenkel, D., and Tolley, G. 1987. “Valuing Changes in Health Risks: A
Comparison of Alternative Measures.” The Southern Economic Journal 53(4):977-984.
Bergman, L. 2005. “CGE Modeling of Environmental Policy and Resource Management.” Handbook of
Environmental Economics 3:1273-1306.
Bovenberg, A. and de Mooij, R. 1994. “Environmental Levies and Distortionary Taxation.” American
Economic Review 84(4): 1085-1089.
Bovenberg, A. and van der Ploeg, F. 1994. “Green Policies and Public Finance in a Small Open Economy.”
Scandinavian Journal of Economics 96(3): 343-363.
Brauer, M., Blair, J., and Vedal, S. 1996. “Effect of Ambient Ozone Exposure on Lung Function in Farm
Workers.” American Journal of Respiratory and Critical Care Medicine 154: 981-987.
Caffet, J-C. 2005. “Health effects and optimal environmental taxes in welfare state countries.” Cahiers
de la Maison des Sciences Economiques 2005.49.
Carbone, J., and Smith, V. 2013. “Valuing Nature in a General Equilibrium.” Journal of Environmental
Economics and Management 66(1): 72-89.
Chambers, C., Chambers, P., and Whitehead, J. 1998. “Contingent Valuation of Quasi-Public Goods:
Validity, Reliability, and Application to Valuing a Historic Site.” Public Finance Review 26(2): 137-154.
Chung-Li, J. 2002. "Including the Feedback of Local Health Improvement in Assessing Costs and Benefits
of GHG Reduction." Review of Urban and Regional Development Studies 14: 282-304.
Crocker, T. and Horst, R. 1981. “Hours of Work, Labor Productivity, and Environmental Conditions: A
Case Study.” The Review of Economics and Statistics 63(3): 361-368.
50
Cropper, M. 2000. “Has economic research answered the needs of environmental policy?” Journal of
Environmental Economics and Management 39(3):328-350.
Cropper, Maureen and Alan J. Krupnick (1999). “The Social Costs of Chronic Heart and Lung Disease,” in
Valuing Environmental Benefits: Selected Essays of Maureen Cropper, eds. Wallace E. Oates and Henk
Folmer, Edward Elgar: Northampton, MA.
Cropper, M., Hammitt, J., Robinson, L., 2011. “Valuing mortality risk reductions: progress and
challenges.” Annual Review of Resource Economics 3:313-336.
Dixon, P., Rimmer, M., and Wittwer, G. 2012. “USAGE-R51, A State-Level Multi-Regional CGE Model of
the U.S. Economy.” GTAP Conference Paper, Purdue University. Available at:
https://www.gtap.agecon.purdue.edu/resources/download/5933.pdf.
Edwards, R. 2008. “Health Risk and Portfolio Choice.” Journal of Business & Economic Statistics 26(4):
472-485.
Evans, W. and Viscusi, W. 1991. “Estimation of State-Dependent Utility Functions Using Survey Data.”
Review of Economics and Statistics 73(1): 94–104.
Fann, N., Lamson, A., Wesson, K., Risley, D., Anenberg, S., and Hubbell, B. 2012. “Estimating the National
Public Health Burden Associated with Exposure to Ambient PM2.5 and Ozone. Risk Analysis.” Risk
Analysis 32(1): 81-95.
Finkelstein, A., Luttmer, E., Notowidigdo, M. 2009. “Approaches to estimating the health state
dependence of the utility function.” American Economic Review Papers and Proceedings 99(2): 116-121.
Finkelstein, A., Luttmer, E., Notowidigdo, M. 2013. “What good is wealth without health? The effect of
health on the marginal utility of consumption.” Journal of the European Economic Association 11(S1):
221–258.
Finnoff, D. and Tschirhart, J. 2008. “Linking Dynamic Economic and Ecological General Equilibrium
Models.” Resource and Energy Economics 30(2): 91-114.
Gayer, T., Hamilton, J., Viscusi, W. K. 2002. “The Market Value of Reducing Cancer Risk: Hedonic Housing
Prices with Changing Information.” Southern Economic Journal 69(2):266-289.
Gerking, S., and Dickie, M. 2013. “Valuing Reductions in Environmental Risks to Children’s Health.”
Annual Review of Resource Economics 5:245-260.
Gerking, S., Dickie, M., and Veronesi, M. 2014. “Valuation of Human Health: An Integrated Model of
Willingness to Pay for Mortality and Morbidity Risk Reductions.” Journal of Environmental Economics
and Management 68(1): 20-45.
Goettle, R., Ho, M., Jorgenson, D., Slesnick, D., and Wilcoxen, P. 2009. “Analyzing Environmental Policies
with IGEM, and Intertemporal General Equilibrium Model of U.S. Growth and the Environment.” Dale
Jorgenson Associates. Available at: http://www.igem.insightworks.com/docs/190.html.
51
Goulder, L. 1995. “Effects of Carbon Taxes in an Economy with Prior Tax Distortions: An Intertemporal
General Equilibrium Analysis.” Journal of Environmental Economics and Management 29(3): 271-297.
Goulder, L., Parry, I., Williams, R., and Burtraw, D. 1999. “The Cost-Effectiveness of Alternative
Instruments for Environmental Protection in a Second-Best Setting.” Journal of Public Economics 72(3):
329-360.
Goulder, L. and Williams III, R. 2003. “The Substantial Bias from Ignoring General Equilibrium Effects in
Estimating Excess Burden, and a Practical Solution.” Journal of Political Economy 111(4): 898–927.
Grosclaude, P. and Soguel, N. 1994. “Valuing Damage to Historic Buildings Using a Contingent Market: A
Case Study of Road Traffic Externalities.” Journal of Environmental Planning and Management 37: 279-
287.
Holland, M., Berry, J., and Forster, D. 1998. ExternE: Externalities of Energy, Methodology 1998 Update,
Volume 7. European Commission, Community Research.
Harrington, W. and Portney, P. 1987. “Valuing the Benefits of Health and Safety Regulation.” Journal of
Urban Economics 22(1):101-112.
Hofstetter, P., Bare, J., Hammitt, J., Murphy, P., and Rice, G. 2002. “Tools for Comparative Analysis of
Alternatives: Competing or Complementary Perspectives?” Risk Analysis 22(5): 833-851.
Hoppe, P., Praml, G., Rabe, G., Lindner, J., Fruhmann, G., and Kessel, R. 1995. “Environmental Ozone
Field Study on Pulmonary and Subjective Responses of Assumed Risk Groups.” Environmental Research
71(2): 109-121.
Johansson, P-O. 1995. Evaluating Health Risks: An Economic Approach. Cambridge, UK: Cambridge
University Press.
Kleckner, N., and J. Neumann. June 3, 1999. “Recommended Approach to Adjusting WTP Estimates to
Reflect Changes in Real Income.” Memorandum to Jim Democker, U.S. EPA/OPAR.
Kling, R., Revier, C., and Sable, K. 2004. “Estimating the Public Good Value of Preserving a Local Historic
Landmark: The Role of Non-substitutability and Citizen Information.” Urban Studies 41(10): 2025-2041.
Korrick, S., Neas, L., Dockery, D., Gold, D., Allen, G., Hill, L., Kimball, K., Rosner, B., and Speizer, F., 1998.
“Effects of Ozone and Other Pollutants on the Pulmonary Function of Adult Hikers.” Environmental
Health Perspectives 106(2): 93-99.
Kuminoff, N., Smith, V., and Timmins, C. 2010. “The New Economics of Equilibrium Sorting and its
Transformational Role for Policy Evaluation.” NBER Working Paper #16349
http://www.nber.org/papers/w16349.
Levy, J., Baxter, L., and Schwartz, J. 2009. “Uncertainty and variability in health-related damages from
coal-fired power plants in the United States.” Risk Analysis 29(7): 1000-1014.
52
Lillard, L, and Weiss, Y. 1997. “Uncertain Health and Survival: Effects on End-of-Life Consumption.”
Journal of Business and Economic Statistics 15(2): 254–268.
Linnerooth, J. 1979. “The value of human life: a review of the models.” Economic Inquiry 17(1):52-74.
Lofgren, H., Harris, R., and Robinson, S. 2001. “A Standard Computable General Equilibrium (CGE) Model
in GAMS.” International Food Policy Research Institute Discussion Paper No. 75.
Matus, K., Yang, T., Paltsev, S., Reilly, J., and Nam, K.-M. 2008. “Toward Integrated Assessment of
Environmental Change: Air Pollution Health Effects in the USA.” Climatic Change 88(1):59-92.
Matus, K., Nam, K.-M., Selin, N., Lamsal, L., Reilly, J., and Paltsev, S. 2012. “Health Damages from Air
Pollution in China.” Global Environmental Change 22(1): 55-66.
Mayeres, I. and van Regemorter, D. 2008. “Modeling the Health Related Benefits of Environmental
Policies and Their Feedback Effects: A CGE Analysis for the EU Countries with GEM-E3.” The Energy
Journal 29(1): 135-150.
Mishan, E. 1971. “Evaluation of life and limb: a theoretical approach.” Journal of Political Economy
79(4):687-705.
Murphy, K., Topel, R. 2006. “The value of health and longevity.” Journal of Political Economy 114(5):
871-904.
Nam, K.-M., Selin, N., Reilly, J., and Paltsev, S. 2010. “Measuring Welfare Loss Caused by Air Pollution in
Europe: A CGE Analysis.” Energy Policy 38: 5059-5071.
OMB. 2003. “Regulatory Analysis: A Primer.” Circular A-4. Washington, DC. Available at
http://www.whitehouse.gov/omb/circulars/a004/a-4.html.
Palmquist, R. 2005. “Property Value Models.” Handbook of Environmental Economics 2:763-819.
Parry, I. 1994. “Pollution Taxes and Revenue Recycling.” Journal of Environmental Economics and
Management 29(3): 64-77.
Parry, I. 1997. “Environmental Taxes and Quotas in the Presence of Distorting Taxes in Factor Markets.”
Resource and Energy Economics 19(3): 203-220.
Parry, I. and Bento, A. 2000. “Tax Deductions, Environmental Policy, and the `Double Dividend’
Hypothesis.” Journal of Environmental Economics and Management 39(1): 67-96.
Parry, I., Williams, R., and Goulder, L. 1999. “When Can Carbon Abatement Policies Increase Welfare?
The Fundamental Role of Distorted Factor Markets.” Journal of Environmental Economics and
Management 37(1): 52-84.
Phaneuf, D., Smith, V. K. 2005. “Recreation Demand Models.” Handbook of Environmental Economics
2:671-761.
Rosen, S. 1988. “The value of changes in life expectancy.” Journal of Risk and Uncertainty 1(3):285-304.
53
Russell, M.W., D.M. Huse, S. Drowns, E.C. Hamel, and S.C. Hartz. 1998. “Direct Medical Costs of Coronary
Artery Disease in the United States.” American Journal of Cardiology 81(9):1110-1115.
Saha, M. and Eckelman, M. 2014. “Urban Scale Mapping of Concrete Degradation from Projected
Climate Change.” Urban Climate 9:101-114.
Samuelson, P. 1974, “Complementarity: An Essay on the 40th Anniversary of the Hicks–Allen Revolution
in Demand Theory.” Journal of Economic Literature 12(4):1255–1289.
Schelling, T. 1968. “The life you save may be your own.” In Chase, S. Jr, (Ed.). Problems in Public
Expenditure Analysis. Washington, DC: The Brookings Institution.
Selin, N., Wu, S., Nam K.-M., Reilly, J., Palsev, S., Prinn, R. 2009. “Global Health and Economic Impacts of
Future Ozone Pollution.” Environmental Research Letters 4: 1-9.
Sieg, H., Smith, V., Banzhaf, S., and Walsh, R. 2004. “Estimating the General Equilibrium Benefits of Large
Changes in Spatially Delineated Public Goods.” International Economic Review 45(4): 1047-1077.
Sloan, F., Viscusi, W., Chesson, H., Conover, C., Whetten-Goldstein, K. 1998. “Alternative approaches to
valuing intangible health losses: the evidence for multiple sclerosis.” Journal of Health Economics
17(4):475-497.
Smith, V. K. 2012. “In search of crosswalks between macroeconomics and environmental economics.”
Review of Environmental Economics and Policy 6(2):298-317.
Smith, V. K., Sieg, H., Banzhaf, H. S., Walsh, R. 2004. “General Equilibrium Benefits for Environmental
Improvements Projected Ozone Reductions Under EPA’s Prospective Analysis for the Los Angeles Air
Basin.” Journal of Environmental Economics and Management 47(2004):559-584.
Tagaris, E., Liao, K., Delucia, A. et al. 2009. “Potential impact of climate change on air-pollution related
human health effects.” Environmental Science and Technology 43(3): 4979–4988.
U.S. EPA. 1999. “The Benefits and Costs of the Clean Air Act 1990 to 2010.” Available at:
http://www.epa.gov/cleanairactbenefits/1990-2010/chap1130.pdf.
U.S. EPA. 2003. Children’s Health Valuation Handbook. EPA 100-R-03-003. Available at:
http://yosemite.epa.gov/ee/epa/eed.nsf/3cdbd09d7c867d9785256c9200548b12/6ed3736d44c87a4a85
256dc1004da4ac/$FILE/handbook1030.pdf.
U.S. EPA. 2004a. “Final Regulatory Analysis: Control of Emissions from Nonroad Diesel Engines.”
Available at: http://www.epa.gov/otaq/documents/nonroad-diesel/420r04007.pdf.
U.S. EPA. 2004b “Review of the Revised Analytical Plan for EPA’s Second Prospective Analysis-Benefits
and Costs of the Clean Air Act 1990-2020: An Advisory by a Special Panel of the Advisory Council on
Clean Air Compliance Analysis.” EPA-SAB-COUNCIL-ADV-04-004. Available at:
http://yosemite.epa.gov/sab/sabproduct.nsf/7CCBBFE15CD4C8B185256F17005E3079/$File/council_adv
_04004.pdf.
54
U.S. EPA. 2007. “Review of the National Ambient Air Quality Standards for Ozone: Policy Assessment of
Scientific and Technical Information.” EPA-452/R-07-007. Available at:
http://www.epa.gov/ttn/naaqs/standards/ozone/data/2007_07_ozone_staff_paper.pdf.
U.S. EPA. 2008. “EMPAX-CGE Model Documentation.” Available at:
http://www.epa.gov/ttnecas1/models/empax_model_documentation.pdf.
U.S. EPA. 2010a. “Valuing Mortality Risk Reductions for Environmental Policy: A White Paper.” Prepared
for review by the Science Advisory Board’s Environmental Economics Advisory Committee. Available at:
http://yosemite.epa.gov/ee/epa/eerm.nsf/vwAN/EE-0563-1.pdf/$file/EE-0563-1.pdf.
U.S. EPA. 2010b. Guidelines for Preparing Economic Analyses. EPA-240-R-10-01. Available at:
http://yosemite.epa.gov/ee/epa/eerm.nsf/vwAN/EE-0568-50.pdf/$file/EE-0568-50.pdf.
U.S. EPA. 2011. The Benefits and Costs of the Clean Air Act 1990 to 2020: EPA Report to Congress.
Washington, DC. Available at http://www.epa.gov/air/sect812/feb11/fullreport_rev_a.pdf.
U.S. EPA. 2012. “Regulatory Impact Analysis (RIA) for the Final Revisions to the National Ambient Air
Quality Standards for Particulate Matter.” Office of Air Quality Planning and Standards, Research
Triangle Park, NC. EPA-452/R-12-005 Available at
http://www.epa.gov/ttn/ecas/regdata/RIAs/finalria.pdf.
U.S. EPA. 2013a. America’s Children and the Environment, Third edition. EPA 240-R-13-001. Available at:
http://www.epa.gov/envirohealth/children/pdfs/ACE3_2013.pdf.
U.S. EPA. 2013b. “Integrated Science Assessment for Ozone and Related Photochemical Oxidants.” EPA
600/R-10/076F. Available at: http://ofmpub.epa.gov/eims/eimscomm.getfile?p_download_id=511347.
U.S. EPA Science Advisory Board. 1998. “Advisory Council on Clean Air Compliance Analysis Advisory on
the Clean Air Act Amendments (CAAA) of 1990 Section 812 Prospective Study: Overview of Air Quality
and Emissions Estimates: Modeling, Health and Ecological Valuation Issues Initial Studies.” EPA-SAB-
COUNCIL-ADV-98-003
Vennemo, H. 1997. “A Dynamic Applied General Equilibrium Model with Environmental Feedbacks.”
Economic Modeling 14(1): 99-154.
Viscusi, W. and Evans, W. 1990. “Utility Functions That Depend on Health Status: Estimates and
Economic Implications.” American Economic Review 80(3): 353–74.
Williams, R. 2002. “Environmental Tax Interactions when Pollution Affects Health or Productivity.”
Journal of Environmental Economics and Management 44(2): 261-270.
Willis, K. 1994. “Paying for Heritage: What Price for Durham Cathedral?” Journal of Environmental
Planning and Management 37(3): 267-278.
Wittels, E.H., J.W. Hay, and A.M. Gotto, Jr. 1990. “Medical Costs of Coronary Artery Disease in the
United States.” American Journal of Cardiology 65(7):432-440.
55
Xu, J., Kochanek, K., Murphy, S., Tejada-Vera, B. 2010. “Deaths: Final data for 2007.” National Vital
Statistics Reports 58(19).
Yamagami, H. 2009. “Health Effects, Tax Deductions for Medical Expenditure, and the Double Dividend.”
FinanzArchiv / Public Finance Analysis 65(1): 105-121.
Yoo, S-H., Kwak, S-J., Lee, J-S. 2008. “Using a Choice Experiment to Measure the Environmental Costs of
Air Pollution Impacts in Seoul.” Journal of Environmental Management 86: 308-318.
Zivin, G. and Neidell, M. 2012. “The Impact of Pollution on Worker Productivity.” American Economic
Review 102(7): 3652-3673.